The Market Encyclopedia
The Market Encyclopedia by 5 Circles Pvt Ltd (SEBI RA INH000020004): a free, open reference on markets — a 400-year timeline, the Hall of Fame and the Hall of Pain, 149 published formulas worked on Indian numbers, algo firms and hubs, exams and books, frauds with their legal status and an honest reality check on retail. English and Hinglish. Education only: no tips, no targets, no recommendations.
5 Circles Private Limited · SEBI Registered Research Analyst · Reg. No. INH000020004 · Educational content only, not investment advice, not a recommendation. Data last verified: 2026-09-07
M1 · Timeline — 400 Years
How markets and each method were born, world and India in parallel.
- 1602 · Dutch East India Company (VOC) issues tradable shares: The VOC was chartered in Amsterdam and raised capital from the public by selling shares that could be resold to others. That secondary trading is why the Amsterdam Bourse is called the first modern stock exchange. Everything after this — brokers, speculation, even short selling — starts here.
- 1730 · Dojima Rice Exchange officially recognised; Homma era: Osaka's Dojima market traded rice receipts and is widely treated as the first organised futures market. Munehisa Homma, an 18th-century rice merchant, is credited in popular history with early candlestick-style price records and crowd-psychology rules. Treat the Homma-candlestick link as tradition, not documented fact.
- 1792 · Buttonwood Agreement — birth of the NYSE: Twenty-four brokers signed a short agreement under a buttonwood tree on Wall Street, fixing commissions and agreeing to trade only among themselves. That club became the New York Stock Exchange. Note the pattern: exchanges begin as private cartels of brokers, and regulation comes much later.
- 1884 · Charles Dow publishes the first stock average: Dow's Customer's Afternoon Letter began carrying an average of eleven stocks, mostly railroads. Averaging many prices into one number was a new idea — it let people talk about 'the market' instead of individual stocks. The Wall Street Journal followed in 1889.
- 1896 · Dow Jones Industrial Average launched: The DJIA started with twelve industrial companies and a simple price average. Being price-weighted, a high-priced stock moves it more than a large company — a flaw that still exists today. It remains the most quoted number in financial news despite better-constructed indices.
- 1902 · Dow Theory takes shape from Dow's editorials: Charles Dow never wrote a book; his Wall Street Journal editorials (1900-1902) were later organised by William Hamilton and Robert Rhea (1932) into 'Dow Theory': trends have three phases, averages must confirm each other, volume confirms trend, a trend persists until reversal is proven.
- 1907 · Livermore era — the 1907 panic short: Jesse Livermore made his name shorting into the Panic of 1907 and again in 1929. His life, told in 'Reminiscences of a Stock Operator' (1923), is the founding myth of the discretionary trader. The honest ending: repeated bankruptcies and his death in 1940. Study the rules; do not copy the life.
- 1929 · The 1929 crash: After a leveraged bull run, the Dow fell about 13% on 28 October and about 12% on 29 October 1929. Margin loans were called, forcing more selling. The decline continued into 1932. The direct results were the Securities Acts and the SEC — modern regulation is a child of this crash.
- 1934 · Graham & Dodd publish 'Security Analysis': Benjamin Graham and David Dodd's textbook separated investing (analysis, safety of principal, adequate return) from speculation. Margin of safety, intrinsic value and reading the balance sheet before the price chart all come from here. Buffett was Graham's student.
- 1934 · US Securities Exchange Act creates the SEC: The Securities Act of 1933 forced disclosure at issue; the 1934 Act created the SEC to police secondary markets, manipulation and insider dealing. The model — a statutory regulator with disclosure at its core — is what India copied in 1992 with SEBI.
- 1952 · Markowitz — 'Portfolio Selection': Harry Markowitz's Journal of Finance paper showed that risk belongs to the portfolio, not the stock: what matters is how holdings move together. Diversification became mathematics — the efficient frontier. Nobel Prize 1990.
- 1962 · Ed Thorp — 'Beat the Dealer': A mathematics professor proved blackjack could be beaten by card counting and bet sizing (Kelly criterion). He then applied the same thinking to warrants and options in 'Beat the Market' (1967) and ran Princeton Newport Partners — the first true quant fund.
- 1964 · Sharpe — CAPM: William Sharpe's Journal of Finance paper gave a price for risk: expected return = risk-free rate + beta × market premium. Beta separated market risk from stock-specific risk. The Sharpe ratio, a way to compare return per unit of risk, came from the same mind. Nobel 1990.
- 1965 · Fama — random walks and efficient markets: Eugene Fama's 1965 work on random walks and his 1970 review 'Efficient Capital Markets' argued that prices already reflect available information, so consistently beating the market is hard. Weak/semi-strong/strong forms come from the 1970 paper. Nobel 2013.
- 1971 · NASDAQ — first electronic quotation system: NASDAQ began as an electronic screen showing dealer quotes, replacing phone calls and pink sheets. It was not yet an order-matching exchange, but it started the shift from floor to screen that NSE would complete in India in 1994.
- 1973 · Black-Scholes-Merton and the CBOE open in the same year: Black and Scholes published their option pricing paper in the Journal of Political Economy (1973); Merton extended it. The Chicago Board Options Exchange opened on 26 April 1973 trading listed calls. A formula and a marketplace arriving together created modern options trading. Nobel 1997 (Scholes, Merton).
- 1982 · First stock index futures (KCBT Value Line, CME S&P 500): Kansas City Board of Trade listed Value Line index futures in February 1982; the CME's S&P 500 contract followed in April. For the first time you could trade 'the market' in one contract, cash-settled. This is the ancestor of Nifty futures.
- 1982 · Renaissance Technologies founded: Mathematician Jim Simons founded Renaissance; the Medallion Fund (1988) became the most successful quant fund on record, staffed by physicists and signal-processing experts rather than finance people. Its returns are known only from press reporting, and the fund is closed to outsiders.
- 1987 · Black Monday — DJIA falls 22.6% in a day: The largest one-day percentage fall in Dow history. The Brady Commission blamed 'portfolio insurance' — automated selling of index futures as prices fell — for accelerating the crash. Circuit breakers were introduced after this. Mechanical strategies that all sell at once create the crash they fear.
- 2001 · US markets complete decimalisation: Stock prices moved from fractions (1/16ths) to cents. Spreads collapsed, which was good for investors but destroyed the old market-maker economics and opened the door to computers that could profit from tiny, fast edges. HFT grows from this soil.
- 2005 · Regulation NMS: The SEC's Regulation NMS required orders to be routed to whichever venue showed the best price. Intended to protect investors, it fragmented US trading across many exchanges and made speed between venues a tradable edge — the world 'Flash Boys' later described.
- 2009 · High-frequency trading becomes the majority of US equity volume: By the late 2000s, firms trading in microseconds accounted for over half of US equity volume, according to widely cited industry estimates. Co-location, direct feeds and rebates made latency a business. Retail cannot compete on speed; it must compete on holding period.
- 2008 · Lehman Brothers files — the Global Financial Crisis: Lehman's bankruptcy on 15 September 2008 turned a US housing/mortgage problem into a global freeze. Models had priced mortgage bonds using correlation assumptions that failed together. Nifty fell about 60% from its January 2008 peak to the October 2008 low.
- 2010 · The Flash Crash: In about 36 minutes the Dow fell roughly 1,000 points and recovered most of it. The SEC/CFTC report pointed to a large automated sell order in E-mini futures interacting with HFT liquidity that vanished. Some stocks printed at one cent. Stop-loss orders became market orders into nothing.
- 2014 · 'Flash Boys' and IEX: Michael Lewis's book told the public how latency arbitrage worked and introduced IEX, an exchange with a deliberate 'speed bump'. The book is journalism, not research, and some claims are contested, but it changed the public conversation about market structure.
- 2012 · Machine learning enters mainstream quant finance: Around the early 2010s, cheaper compute, open-source libraries (scikit-learn, later TensorFlow/PyTorch) and alternative data brought ML from a few secretive funds to every quant desk. The failure mode also went mainstream: overfitting to noise, and backtests that look perfect until traded.
- 2018 · 'Advances in Financial Machine Learning' (López de Prado): Marcos López de Prado's book (Wiley, 2018) gave practitioners tools for the specific ways finance breaks ML: purged cross-validation, the deflated Sharpe ratio, probability of backtest overfitting, and meta-labelling. It is the standard reference for honest backtesting.
- 2020 · COVID crash — fastest bear market on record: Global equities fell over 30% in about a month. US circuit breakers tripped four times in March 2020; Nifty fell about 13% on 23 March 2020, its worst single day. The rebound was equally fast, driven by central-bank liquidity — and it minted a generation of new retail traders in India.
- 2021 · GameStop short squeeze: Retail traders coordinating on Reddit pushed GameStop from under $20 to an intraday high near $483 on 28 January 2021, squeezing hedge funds that were short more than 100% of the float. Brokers restricted buying the same day, citing clearing-house margin. Many latecomers bought the top.
- 2023 · 0DTE options boom in the US: After Cboe added Tuesday and Thursday SPX expiries in 2022, every weekday had a same-day expiry. Cboe's own data has reported zero-days-to-expiry contracts at roughly 40-50% of SPX options volume in 2023-2025. India's weekly index options reached the same place first — SEBI's FY25 data shows 70% of index-option turnover on expiry day.
- 1855 · Brokers gather under a banyan tree in Bombay: From the 1850s a handful of brokers met under a banyan tree near Town Hall, Bombay, trading shares of banks and cotton presses. The American Civil War cotton boom of the 1860s swelled their numbers and produced India's first share mania and bust (1865).
- 1875 · The Native Share & Stock Brokers' Association — today's BSE: The brokers formalised their club as The Native Share & Stock Brokers' Association, Asia's oldest stock exchange. It moved to Dalal Street and became the Bombay Stock Exchange. Like Buttonwood, it began as a self-regulating broker cartel with entry fees and its own rules.
- 1900 · Badla — India's homegrown carry-forward system: Badla let a buyer carry a position past settlement by paying a financing charge (badla rate) to a financier, with 'undha badla' for short sellers. It was leverage without margin discipline and was suspended and revived repeatedly. It was finally banned in July 2001 after the Ketan Parekh episode.
- 1992 · Harshad Mehta securities scam exposed: Mehta used fake bank receipts and the ready-forward market to divert bank money into stocks, running ACC and others up many times. Sucheta Dalal's Times of India report in April 1992 exposed it; the estimated diversion was around ₹4,000-5,000 crore (1992 rupees). The Sensex fell sharply and Parliament's JPC followed.
- 1992 · SEBI Act 1992 — statutory powers: SEBI had existed since 1988 as a non-statutory body. The SEBI Act 1992 gave it statutory powers to register and regulate intermediaries, investigate, and penalise. The Controller of Capital Issues was abolished the same year, freeing companies to price their own issues.
- 1994 · NSE begins screen-based trading: NSE, incorporated in 1992, started wholesale debt trading in June 1994 and equities in November 1994 on a fully electronic, order-driven, nationwide system. Within a year it overtook BSE's open-outcry floor. Anonymous, price-time priority matching ended the broker's information edge.
- 1996 · Depositories Act; NSDL starts dematerialisation: Paper share certificates meant forgery, bad deliveries and months-long transfers. The Depositories Act 1996 and NSDL (promoted by NSE and others) put shares in electronic form. CDSL, promoted by BSE, followed in 1999. Your demat account exists because of this.
- 1999 · CDSL begins operations: Central Depository Services (India) Ltd, promoted by BSE, started in 1999 as the second depository. Competition between NSDL and CDSL cut demat costs; CDSL later became the depository with the most retail accounts as discount brokers grew after 2020.
- 2000 · Nifty index futures launched: NSE and BSE launched index futures on 12 June 2000 — India's first exchange-traded equity derivative, following the L.C. Gupta committee and the Securities Laws (Amendment) Act 1999 that made derivatives legal 'securities'. Volumes were tiny for the first year.
- 2001 · Ketan Parekh scam and the K-10 stocks: Parekh ran up a set of tech/media stocks ('K-10') using funds routed through Madhavpura Mercantile Co-operative Bank and pay-orders. The unwind in March 2001, around the Union Budget, crashed prices and broke the bank. SEBI banned him from markets; the episode triggered the badla ban and rolling settlement.
- 2001 · Nifty index options launched: European-style index options on Nifty began on 4 June 2001. Two decades later, index options became the single largest derivative product on earth by contracts, and the source of roughly 92% of Indian individual traders' F&O losses (SEBI FY26 study).
- 2001 · Badla banned; rolling settlement for all stocks: From 2 July 2001 SEBI banned all deferral products (badla/ALBM) and moved actively traded stocks to compulsory rolling settlement, initially T+5, cut to T+3 in 2002 and T+2 in April 2003. Leverage moved from the informal badla market to the regulated F&O segment.
- 2001 · Options on individual stocks launched: Stock options (American-style at the time) on a first list of 31 stocks began on 2 July 2001 — the same day badla ended. Stock options were later made European-style (2011). Liquidity has always concentrated in a few dozen names.
- 2001 · Single-stock futures launched: India launched futures on individual stocks on 9 November 2001, a product few large markets offered. They quickly became the most traded stock derivative in India, effectively the regulated successor to badla's leveraged carry-forward.
- 2008 · SEBI permits Direct Market Access — algos arrive: SEBI's April 2008 circular allowed brokers to give institutional clients direct electronic access to exchange systems without manual intervention. This is the legal birth of algorithmic trading in India; it was extended step by step until retail API trading became common after 2015.
- 2009 · Satyam — chairman confesses to fabricated accounts: Ramalinga Raju admitted that Satyam's cash and bank balances had been inflated by thousands of crores over years, with auditors signing off. The stock fell about 78% in a day. India's biggest corporate accounting fraud led to the Companies Act 2013 governance rules and NFRA. Raju was convicted in 2015.
- 2010 · NSE offers co-location: NSE began renting rack space in its data centre so members' servers could sit next to the matching engine, cutting latency to microseconds. Co-location is standard worldwide, but how fairly its data feeds were distributed became the subject of India's biggest market-structure scandal.
- 2013 · NSEL payment crisis: National Spot Exchange Ltd, a 'spot' exchange, was in practice offering paired contracts that worked like financing at fixed returns, backed by warehouse stock that often did not exist. In July 2013 it defaulted on about ₹5,600 crore owed to roughly 13,000 investors. Recovery litigation continues.
- 2019 · SEBI order in the NSE co-location matter: SEBI found that NSE's tick-by-tick data architecture allowed some members preferential access, ordered NSE to disgorge about ₹625 crore plus interest, and barred it from the securities market for six months. In January 2023 the Securities Appellate Tribunal set aside the disgorgement and instead directed NSE to pay ₹100 crore for lack of due diligence. SEBI appealed to the Supreme Court; in 2026 NSE settled the co-location and dark fibre cases with SEBI for ₹1,491.21 crore, and in September 2026 the Supreme Court disposed of SEBI's appeals against NSE. Status: SETTLED (no admission).
- 2022 · SEBI order on Chitra Ramkrishna and the 'Himalayan yogi': SEBI's February 2022 order found that NSE's former CEO shared confidential exchange information with an unknown outsider and that governance failures allowed Anand Subramanian's irregular appointment. Penalties were imposed on Ramkrishna (₹3 crore) and on NSE, Ravi Narain and Anand Subramanian (₹2 crore each). Separately, the CBI arrested her in March 2022 in the co-location case; she was granted bail in September 2022. Status: UNDER APPEAL (SAT admitted the appeals in 2022; no final ruling reported).
- 2023 · India completes the move to T+1 settlement: Phased in from February 2022 by market-cap bucket, the last batch of stocks moved to T+1 on 27 January 2023, making India the first major market to settle the day after trade — ahead of the US (May 2024). Faster settlement means less counterparty risk and less margin locked.
- 2023 · SEBI's first F&O loss study: 89% of individuals lost in FY22: SEBI published its first population-scale study of individual traders in equity F&O: 89% lost money in FY22, average loss about ₹1.1 lakh, and the number of individual traders had grown many times since FY19. It reframed the F&O boom as a consumer-protection problem.
- 2024 · T+0 settlement beta launched: SEBI allowed an optional same-day (T+0) settlement cycle for a small set of stocks from 28 March 2024, alongside T+1. It was later widened to more stocks and brokers. Adoption has been slow because the T+1 cycle already works well for most participants.
- 2024 · SEBI: 93% lost over FY22-24, aggregate loss over ₹1.8 lakh crore: SEBI's three-year study found that 93% of individual F&O traders lost money across FY22-FY24, with aggregate net losses above ₹1.8 lakh crore and average losses around ₹2 lakh per person over the period. Proprietary desks and FPIs were the gross winners. Weekly-expiry curbs and higher lot sizes followed within weeks.
- 2025 · SEBI circular: safer participation of retail investors in algo trading: Circular SEBI/HO/MIRSD/MIRSD-PoD/P/CIR/2025/0000013 set the retail algo framework: static-IP whitelisting for API orders, a 10 orders-per-second threshold above which strategies need exchange registration, unique strategy IDs on every order, broker-routed only, and a Research Analyst licence for black-box algo providers. Deadline extended by circular 2025/132 to 1 April 2026.
- 2025 · SEBI: 91% lost in FY25; net loss ₹1,05,603 crore: The FY25 update found 91% of individual equity-derivatives traders lost money and aggregate net losses widened about 41% year on year to ₹1,05,603 crore, despite the curbs introduced in late 2024. Average loss per person was again around ₹1.1 lakh.
- 2025 · SEBI interim order against Jane Street (Bank Nifty): SEBI barred Jane Street Group from Indian markets and impounded ₹4,843.57 crore, alleging manipulation of Bank Nifty on expiry-linked days between January 2023 and March 2025, with further alleged instances in May 2025 after an NSE caution letter. The bar was lifted after the amount was placed in escrow; Jane Street appealed to SAT in September 2025 and the matter remains unresolved. Status: INTERIM ORDER — UNDER APPEAL.
- 2026 · Retail algo framework becomes mandatory: From 1 April 2026 every broker and retail API user must comply: orders only from whitelisted static IPs, OAuth + 2FA with daily session expiry, strategy IDs on every order, 10 OPS threshold for registration, immediate-family-only sharing of self-built algos, and 5-year audit logs at the broker. NSE operationalised it via circular NSE/INVG/67858. This is current law, not an upcoming change.
- 2026 · Union Budget 2026-27 raises STT on F&O: For derivative trades entered on or after 1 April 2026, STT went up on the sale of futures, on the sale of options (on premium) and on exercise. STT rates change with budgets — check the current rate on your contract note. The stated purpose was to moderate speculative derivatives activity.
- 2026 · NSE remains the world's largest derivatives exchange by contracts: FIA's exchange-traded derivatives statistics have ranked NSE first by number of contracts traded every year since 2019, and FIA's 2026 monthly reports say it held that place through the first four months of 2026 — even after SEBI's curbs cut global options volume sharply in 2025. The rank is by contract count; by notional value US exchanges are larger.
- 2026 · SEBI: 87.7% lost in FY26; net loss ₹91,685 crore; 90% of two-year losers lost again: The FY26 study showed the first fall in aggregate losses (₹91,685 crore vs a restated ₹1,11,788 crore in FY25), but only because participation fell — active traders down about 20% to 78.6 lakh, new entrants down about 40%. Average net loss per trader rose to about ₹1.17 lakh. Options drove about 92% of losses. Of those who lost two years running and kept trading, about 90% lost again.
M2 · Hall of Fame
People who made the money, their actual method, and their worst loss.
- Benjamin Graham (USA (born UK)): Father of value investing; taught Buffett margin of safety.
- David Dodd (USA): Graham's co-author; turned lecture notes into Security Analysis.
- Warren Buffett (USA): Turned a textile mill into the world's biggest holding company.
- Charlie Munger (USA): Buffett's partner; moved value investing toward quality businesses.
- Philip Fisher (USA): Scuttlebutt investor; buy great growth companies and almost never sell.
- Sir John Templeton (USA / Bahamas): Global bargain hunter who bought at the point of maximum pessimism.
- Peter Lynch (USA): Ran Magellan at 29% a year; invest in what you know.
- Joel Greenblatt (USA): Special-situations investor who wrote the Magic Formula.
- Seth Klarman (USA): Oracle of Boston; holds cash until a true bargain appears.
- Howard Marks (USA): Distressed-debt pioneer famous for his investor memos.
- Mohnish Pabrai (India / USA): Buffett cloner; heads I win, tails I don't lose much.
- Guy Spier (Switzerland (born South Africa)): Runs Aquamarine Fund; investing as self-improvement.
- Li Lu (USA (born China)): The 'Chinese Buffett'; the only outsider Munger trusted with money.
- Terry Smith (UK): Buy good companies, don't overpay, do nothing.
- Nick Sleep (UK): Nomad letters; scale economics shared; walked away at the top.
- Chuck Akre (USA): Three-legged stool: great business, great people, great reinvestment.
- Bill Ackman (USA): Activist who takes huge concentrated bets and fights in public.
- Aswath Damodaran (USA (born India)): Dean of Valuation; every number must tell a story.
- Joseph Piotroski (USA): Nine yes/no checks that separate cheap winners from cheap losers.
- Edward Altman (USA): Z-Score inventor; predicting bankruptcy from five ratios since 1968.
- Messod Beneish (USA): M-Score: eight ratios that flag likely earnings manipulation.
- James Ohlson (USA (born Sweden)): O-Score and the residual-income model of equity value.
- Richard Sloan (USA): Found the accrual anomaly: paper profits fade, cash profits last.
- Howard Schilit (USA): Sherlock Holmes of accounting; wrote Financial Shenanigans.
- Charles Dow (USA): Built the first stock average and the theory named after him.
- Munehisa Homma (Japan): Sakata rice trader credited as father of candlestick charting.
- Richard Wyckoff (USA): Read the 'composite operator' through price, volume and structure.
- Ralph Nelson Elliott (USA): Accountant who saw markets move in five-and-three wave patterns.
- W.D. Gann (USA): Geometry, time cycles and angles; divisive but still followed.
- Jesse Livermore (USA): Boy Plunger; made and lost fortunes reading the tape.
- Nicolas Darvas (Hungary / USA): Dancer who made $2 million with the box method by telegram.
- Richard Donchian (USA): Father of trend following; the four-week breakout rule.
- J. Welles Wilder Jr. (USA / New Zealand): Engineer who invented RSI, ATR, ADX and Parabolic SAR.
- Gerald Appel (USA): Invented MACD; ran a newsletter and money manager for 50 years.
- John Bollinger (USA): Bollinger Bands: volatility-scaled envelopes around a moving average.
- Stan Weinstein (USA): Stage analysis: buy Stage 2, avoid Stage 4, using the 30-week average.
- William O'Neil (USA): CAN SLIM creator; founded Investor's Business Daily.
- Thomas Bulkowski (USA): Measured thousands of chart patterns; wrote the Encyclopedia.
- Steve Nison (USA): Brought Japanese candlesticks to the West in 1991.
- John J. Murphy (USA): Wrote the standard technical-analysis textbook; father of intermarket analysis.
- Martin Pring (UK / USA): Technical Analysis Explained; business-cycle sector rotation.
- Alexander Elder (USA (born USSR / Estonia)): Psychiatrist-trader; Triple Screen, Force Index, Elder-Ray.
- Mark Minervini (USA): SEPA and the VCP; 1997 US Investing Championship winner.
- David Ryan (USA): O'Neil's protege; three straight US Investing Championship wins.
- Dan Zanger (USA): Pool contractor who ran $10,775 to $18 million in the dot-com boom.
- Linda Bradford Raschke (USA): Short-term swing trader; the Holy Grail and Turtle Soup setups.
- Marty Schwartz (USA): Pit Bull; nine-year losing analyst turned champion S&P trader.
- Peter Brandt (USA): Classical chartist trading futures since 1975; publishes his own drawdowns.
- Tom DeMark (USA): TD Sequential and Combo: counting bars to time exhaustion.
- Perry Kaufman (USA): Trading Systems and Methods; the adaptive moving average.
- Brian Shannon (USA): Multiple timeframes and the anchored VWAP.
- Richard Dennis (USA): Prince of the Pit; taught the Turtles that trading can be learned.
- Curtis Faith (USA): Youngest Turtle; wrote Way of the Turtle.
- Jerry Parker (USA): Most successful Turtle; runs Chesapeake Capital since 1988.
- Ed Seykota (USA): Pioneer of computerised trend systems; everybody gets what they want from markets.
- Bill Dunn (USA): Founded DUNN Capital in 1974; decades of pure trend following.
- John W. Henry (USA): Trend-following CTA fortune bought the Red Sox and Liverpool.
- David Harding (UK): Co-founded AHL and Winton; statistics applied to trend following.
- Michael Covel (USA): Chronicler of trend following; wrote the Turtle story.
- Robert Carver (UK): Ex-AHL quant who open-sourced systematic trading for individuals.
- Rakesh Jhunjhunwala (India): India's most famous bull; bet on India's growth for four decades.
- Radhakishan Damani (India): Quiet value investor who built DMart; Jhunjhunwala's mentor.
- Vijay Kedia (India): SMILE investor: small in size, medium in experience, large in aspiration, extra-large in market potential.
- Ramesh Damani (India): BSE member who bought Infosys early and preaches long holding periods.
- Raamdeo Agrawal (India): QGLP: quality, growth, longevity, price — the Motilal Oswal wealth-creation lens.
- Chandrakant Sampat (India): India's original value investor; bought MNC consumer stocks and never sold.
- Ashish Dhawan (India): Private-equity pioneer turned philanthropist; concentrated public-market positions.
- Ashish Kacholia (India): Low-profile small-cap specialist tracked closely by retail investors.
- Dolly Khanna (India): Chennai-based small-cap investor; portfolio managed by husband Rajiv Khanna.
- Porinju Veliyath (India): Kochi's contrarian: buys hated, misgoverned small caps for a turnaround.
- Saurabh Mukherjea (India): Coffee Can investing: buy consistent compounders, do nothing for a decade.
- Prashant Jain (India): One of India's longest-serving fund managers; contrarian value at HDFC AMC.
- Kenneth Andrade (India): Buys industry leaders at the bottom of their capex cycle.
- Sunil Singhania (India): Ex-Reliance MF CIO; small-cap growth at reasonable price via Abakkus.
- Samir Arora (India): Long-short India fund manager from Singapore; blunt, contrarian, media-savvy.
- Madhusudan Kela (India): Mid-cap and cyclical specialist; former Reliance Capital investment strategist.
- Basant Maheshwari (India): Kolkata growth investor; 'The Thoughtful Investor'; pays up for leaders.
- Shankar Sharma (India): Perma-contrarian; First Global co-founder; now global small-cap quant tilt.
- Nilesh Shah (India): Kotak AMC MD; institution-builder who popularised SIP discipline.
- Parag Parikh (India): Brought behavioural finance to Indian value investing; founded PPFAS.
- Vallabh Bhanshali (India): Enam co-founder; investment banker turned philosophical long-term investor.
- George Soros (Hungary / United States): Reflexivity theorist; the man who broke the Bank of England.
- Stanley Druckenmiller (United States): Soros's chief lieutenant; concentration and liquidity-driven macro.
- Paul Tudor Jones II (United States): Called the 1987 crash; defence-first macro trader.
- Bruce Kovner (United States): Caxton founder; macro with a scholar's research and a cab driver's humility.
- Michael Steinhardt (United States): Contrarian 'variant perception' hedge-fund pioneer with a famous temper.
- Ray Dalio (United States): Built the world's largest hedge fund on systematised macro 'principles'.
- Louis Moore Bacon (United States): Moore Capital founder; profited in the 1987 crash and the 1990 Gulf shock.
- Jim Rogers (United States / Singapore): Soros's co-founder turned commodities bull and world-travelling macro contrarian.
- Julian Robertson (United States): Tiger Management founder; father of the 'Tiger Cubs'.
- Steven A. Cohen (United States): Tape-reading legend behind SAC and Point72; Mets owner.
- Israel 'Izzy' Englander (United States): Built Millennium's multi-manager 'pod' machine from $35 million in 1989.
- Alfred Winslow Jones (Australia / United States): Sociologist-journalist who invented the hedged fund in 1949.
- John Paulson (United States): The 'greatest trade ever': shorting subprime in 2007.
- Michael Burry (United States): Doctor turned value investor who shorted subprime through CDS.
- James Chanos (United States): Career short seller who called Enron before it collapsed.
- David Einhorn (United States): Greenlight founder; the Lehman short and 'Fooling Some of the People'.
- J. Kyle Bass (United States): Subprime and sovereign-debt short seller; Hayman Capital founder.
- Edward O. Thorp (United States): Beat blackjack, then the market: father of quantitative hedge funds.
- James H. Simons (United States): Codebreaker-mathematician whose Medallion Fund is the best record ever.
- Leonard E. Baum (United States): Co-inventor of the Baum–Welch algorithm; Simons's first trading partner.
- Elwyn Berlekamp (United States): Coding theorist who re-engineered Medallion into a short-horizon machine.
- Robert Mercer (United States): IBM speech-recognition scientist who became Renaissance co-CEO.
- Peter Brown (United States): Renaissance CEO; co-architect of Medallion's statistical equity engine.
- David E. Shaw (United States): Computer scientist who built D. E. Shaw and mentored Jeff Bezos.
- John Overdeck (United States): Maths olympian, ex-D. E. Shaw and Amazon; Two Sigma co-founder.
- David Siegel (United States): MIT AI PhD; Two Sigma co-founder focused on technology and infrastructure.
- Clifford Asness (United States): Fama's student who industrialised value and momentum at AQR.
- Emanuel Derman (South Africa / United States): Physicist-quant; co-author of the Black–Derman–Toy model and local volatility.
- Nassim Nicholas Taleb (Lebanon / United States): Options trader who taught the world about fat tails and Black Swans.
- Marcos López de Prado (Spain / United States): Wrote the manual for machine learning in finance and against backtest overfitting.
- Harry Markowitz (United States): Modern Portfolio Theory: the first maths of diversification. Nobel 1990.
- William F. Sharpe (United States): CAPM and the Sharpe ratio; Nobel 1990.
- Eugene F. Fama (United States): Efficient Market Hypothesis and the Fama–French factors; Nobel 2013.
- Kenneth R. French (United States): Fama's co-author; keeper of the factor data library.
- Robert C. Merton (United States): Extended Black–Scholes with continuous-time maths; Nobel 1997; LTCM partner.
- Myron Scholes (Canada / United States): Co-author of Black–Scholes; Nobel 1997; LTCM partner.
- Fischer Black (United States): Black–Scholes co-author and Goldman quant; died before the Nobel.
- Robert F. Engle (United States): Invented ARCH: volatility clusters, and can be modelled. Nobel 2003.
- Benoit Mandelbrot (France / United States): Fractal geometry; showed cotton prices have fat tails, not bell curves.
- Andrew W. Lo (United States): Adaptive Markets Hypothesis: efficiency evolves; so must strategies.
- Robert J. Shiller (United States): Irrational Exuberance, the CAPE ratio and the Case–Shiller index; Nobel 2013.
- Richard H. Thaler (United States): Behavioural economics into practice: nudges and mispricing funds. Nobel 2017.
- Daniel Kahneman (Israel / United States): Prospect theory and Thinking, Fast and Slow; Nobel 2002.
- Amos Tversky (Israel / United States): Kahneman's partner; the mathematical mind behind prospect theory.
- Thomas Peterffy (Hungary / United States): Invented electronic market making; founded Interactive Brokers.
- Jeff Yass (United States): Poker-playing options market maker; Susquehanna co-founder.
- Kenneth C. Griffin (United States): Citadel founder; hedge fund and market maker behind one in four US stock trades.
- Vincent Viola (United States): NYMEX chairman turned Virtu founder; the firm that almost never has a losing day.
- Donald R. Wilson Jr. (United States): Chicago options-pit trader who built DRW, an early crypto market maker.
- Mark Gorton (United States): Founded Tower Research Capital and LimeWire; Tower settled a spoofing case in 2019.
- Alexander Gerko (Russia / United Kingdom): Built XTX Markets into one of the world's largest FX and equity liquidity providers.
- Brad Katsuyama (Canada / United States): Flash Boys protagonist; built IEX and its 'speed bump' against latency arbitrage.
- Sandeep Tyagi (India / United States): Estee Advisors founder; among the first SEBI-registered quant/algo shops in India.
M3 · Hall of Pain
Blow-ups, and the exact mechanism that broke.
- Long-Term Capital Management (1998, USA): Russia defaulted in Aug 1998, spreads blew out everywhere at once, and the fund could not meet margin. In Sept 1998 fourteen banks and brokerage firms, brought together by the New York Fed, put $3.6 bn into the fund; the Fed lent none of its own money. Your model's worst case is not the market's worst case. Size for the day correlations break.
- Nick Leeson / Barings Bank (1995, UK / Singapore): The Jan 1995 Kobe earthquake sent the Nikkei down; losses ballooned and the 233-year-old bank collapsed. The person taking risk must never be the person checking it. Never average a losing position to hide it.
- Jérôme Kerviel / Société Générale (2008, France): The bank found the positions in Jan 2008 and unwound them into a falling market over three days. Gross exposure matters, not just net. If your 'hedge' is imaginary, so is your risk number.
- Brian Hunter / Amaranth Advisors (2006, USA / Canada): The spread collapsed in Sept 2006; the fund could not exit a position bigger than the market and was liquidated. If you cannot exit in a week without moving the price, you do not own a position, the position owns you.
- Bill Hwang / Archegos Capital (2021, USA): A ViacomCBS share sale in March 2021 broke the price; margin calls hit every broker at once and banks dumped blocks. Credit Suisse alone lost about $5.5 bn. Leverage plus concentration plus illiquidity is the same trade every blow-up makes. Whoever holds the biggest position is the market.
- Gabe Plotkin / Melvin Capital (2021, USA): A retail-driven squeeze in Jan 2021 sent GameStop up many times over; the fund lost 53% in one month (~$6.8 bn) and needed a $2.75 bn injection from Citadel and Point72. It wound down in May 2022. Being right about the business is not the same as being safe in the position. Shorts need a hard stop, not a thesis.
- Knight Capital (2012, USA): Old dormant code was reactivated on one server; the firm accumulated a huge unwanted inventory and was sold within months. Every algo needs a kill switch and a hard order-rate/position cap outside the algo itself. SEBI's retail algo framework (circular of 4 Feb 2025, fully mandatory for brokers from 1 Apr 2026) requires a kill switch for exactly this kind of failure.
- Victor Niederhoffer (1997, USA): The Oct 27, 1997 crash forced liquidation; ten years later the 2007 credit shock closed his next fund. A strategy that blew you up once will do it again unless the sizing rule changed. Selling naked puts is not income, it is insurance you wrote.
- Jesse Livermore (1940, USA): He made and lost fortunes repeatedly, declared bankruptcy in the 1930s, and died by suicide in 1940. Skill without a survival rule is a slow bankruptcy. Keep money the market can never touch.
- Richard Dennis (1988, USA): After the 1987 crash the public funds suffered heavy drawdowns and were closed; Dennis stepped back from managing outside money. Max drawdown is a business constraint, not just a stat. Trade the size your investors can survive, not the size your edge allows.
- MF Global / Jon Corzine (2011, USA): Margin calls and a ratings downgrade drained liquidity; the broker filed for bankruptcy and customer segregated funds were found about $1.6 bn short (the trustee later recovered the money, and customers were repaid in full). Broker risk is real. Client money is the last line that must never be crossed — the Indian Karvy and Anugrah cases are the same story.
- Bear Stearns High-Grade funds (2007, USA): Subprime marks fell in spring 2007, repo lenders pulled funding, and both funds collapsed by July 2007 — the first domino of the crisis. A rating is someone else's opinion about a model. Funding an illiquid asset overnight is a bet on other people's calm.
- Lehman Brothers (2008, USA): Confidence went, funding stopped rolling, no buyer or rescue came, and it filed the largest bankruptcy in US history on 15 Sept 2008. Too big to fail is a hope, not a hedge. Watch the funding side of a balance sheet, not only the asset side.
- Three Arrows Capital (3AC) (2022, Singapore / BVI): LUNA went to zero in May 2022, GBTC discount widened, lenders called loans; the fund entered BVI liquidation in June 2022 owing creditors about $3.3 bn. Counterparty and leverage risk hide in 'uncorrelated' crypto exactly like in bonds. Who lent you money is also who will liquidate you.
- FTX / Alameda Research (2022, Bahamas / USA): A leaked balance sheet (Nov 2022) and a rival's FTT sale triggered a run; FTX filed for bankruptcy within days. Not your keys, not your coins — and not your segregated account, not your money. Custody is the first risk, before any strategy.
- Credit Suisse: Greensill + Archegos (2021, Switzerland): Greensill collapsed in March 2021; Archegos blew up the same month. Credit Suisse lost ~$5.5 bn on Archegos and by mid-2024 investors had got back more than $7 bn of the roughly $10 bn in the Greensill-linked funds, and UBS then offered to pay 90% of what was still owed. The bank was absorbed by UBS in 2023. Institutions blow up the same way individuals do: one client, one product, too much of it. Fee greed is leverage in disguise.
- XIV inverse-VIX ETN termination (2018, USA): On 5 Feb 2018 the VIX jumped a record 115.6% in one day; the note lost more than 90% of its value late that day, and the issuer used its acceleration clause to terminate it (last trading day 20 Feb 2018). Read the termination clause. Products that 'always make money' have a day they are designed to die.
- Negative WTI crude, 20 Apr 2020 (2020, USA / global): With storage space at Cushing running out, holders paid to get rid of delivery obligations; on 20 April 2020 the May contract settled at minus $37.63 a barrel, below zero for the first time ever. Never hold a physically-delivered contract into expiry unless you can take delivery. 'Cannot go lower' is not a risk model.
- Bill Ackman: Herbalife short + Valeant long (2017, USA): The FTC made Herbalife pay $200 million and restructure its business (July 2016) but did not call it a pyramid scheme or shut it; Pershing exited the short in February 2018 at a loss. Valeant collapsed on pricing and accounting questions; Pershing exited in March 2017. Ego is a position size. The more publicly you argue a trade, the harder it is to cut.
- Karvy Stock Broking clients (2019, India): Karvy used the power of attorney to pledge client securities and raise money for group entities; SEBI found Rs 1,443 crore diverted, barred the broker and its promoter for 7 years and imposed Rs 21 crore penalties (April 2023). Check your CDSL/NSDL CAS every month. Use DDPI/e-DIS, not a blanket PoA. Your broker is a counterparty.
- Anugrah Stock & Broking clients (2020, India): SEBI found Anugrah collected funds under assured-return schemes and misused client money, by as much as about Rs 119 crore on the days SEBI sampled; the broker was suspended (Nov 2020), fined, and in February 2023 barred with its promoters for up to 7 years, with Rs 8 crore in penalties. A SEBI registration number is for a specific activity. Assured returns are illegal for every registered category — the promise itself is the red flag.
- BRH Wealth Kreators clients (2019, India): SEBI's interim order against BRH Wealth Kreators (October 2019) came weeks before the larger Karvy matter. Its final order (January 2023) found client funds and securities misappropriated, barred eight entities for up to 7 years and imposed Rs 11.5 crore in penalties. The pattern repeats. Reconcile broker ledger with depository statement, not with the broker's app.
- IL&FS default (2018, India): Group entities began defaulting in Sept 2018 with debt of more than Rs 91,000 crore; the government superseded the board via NCLT and SFIO investigated. About Rs 48,463 crore of a Rs 61,000 crore resolution target had been repaid by Sept 2025. Ratings lag reality. A AAA that becomes D in weeks means the rating was never information. Watch interest coverage and short-term debt yourself.
- DHFL default (2019, India): DHFL defaulted in 2019 and was the first financial company sent to IBC by RBI. CBI booked it in June 2022 for what it called the biggest bank fraud, Rs 34,615 crore, against a 17-bank consortium. Retail NCDs at 1-2% above bank FD are not 'slightly better FDs'. Read the rating rationale and the promoter's other businesses.
- India's retail F&O cohort (2025, India): SEBI's study found 93% of individual traders lost money in equity F&O over FY22-FY24, aggregate net losses above Rs 1.8 lakh crore. In FY25, 91% lost and net losses widened to Rs 1,05,603 crore (SEBI later revised FY25 to about Rs 1.12 lakh crore). In FY26, after SEBI's curbs, 87.7% still lost; net losses eased to Rs 91,685 crore as fewer individuals traded. This is the base rate. Any course, tip or 'strategy' that does not start from this number is selling you the lucky 1 in 10. Link: reality-check module.
- Paytm IPO retail investors (2021, India): The stock closed 27% below issue price on listing day (18 Nov 2021), erasing about Rs 38,000 crore of market value; it later traded 70% below the IPO price. Zomato, by contrast, listed at a premium. An IPO is the one time the seller knows more than you and picks the price. Loss-making + grey-market hype = skip.
- Yes Bank AT1 bondholders (2020, India): In the March 2020 rescue, AT1 bonds worth Rs 8,415 crore were written to zero while equity survived. Bombay HC struck the write-down down in January 2023; on 15 October 2026 the Supreme Court dismissed the appeals of Yes Bank, the RBI and the Centre and upheld that ruling. Yes Bank was reported to be preparing a review petition. Yield is the price of a clause you did not read. If a 'bond' can legally go to zero before the shares, it is not a bond for you.
- PMC Bank depositors (2019, India): RBI imposed withdrawal limits in Sept 2019 after the bank was found to have hidden a huge exposure to one borrower group (HDIL); the EOW FIR put the wrongful loss at Rs 4,355 crore. Depositors waited years for staggered access. Deposit insurance has a cap and co-operative banks have weaker oversight. Spread deposits; never keep emergency money in one small bank.
- Gensol Engineering shareholders (2025, India): SEBI's interim order (15 Apr 2025) alleged loan funds meant for EVs were diverted for promoters' personal use, including luxury property; the stock fell about 90% from its high, promoters resigned, the company entered insolvency, and SEBI confirmed the interim directions in July 2025 (its findings are still prima facie, pending a forensic audit). Story stocks with rising debt and related-party deals: check where the borrowed money went before checking the order book.
M4 · The Formula Vault
Every published formula, explained and worked on Indian numbers.
- DCF — Free Cash Flow to Firm (FCFF): Adds up every future rupee of cash the business will throw off for all capital providers, discounted back to today at the firm's WACC. Subtract net debt to get equity value.
EV = sum_{t=1..n} FCFF_t / (1+WACC)^t + [FCFF_n x (1+g) / (WACC - g)] / (1+WACC)^n ; Equity = EV - Net debtFails when (read this): Banks/NBFCs (working capital and debt are the business itself), loss-making start-ups, companies with lumpy capex cycles, and any time terminal value is >75% of the answer. - DCF — Free Cash Flow to Equity (FCFE): Values only the shareholders' slice: cash left after capex, working capital AND net debt repayment, discounted at cost of equity (not WACC).
FCFE = Net income + D&A - Capex - dWC + (New debt - Debt repaid); Equity value = sum FCFE_t/(1+ke)^t + TVFails when (read this): Negative FCFE for years (heavy-capex build-outs), firms that borrow to pay dividends, financial companies. - Gordon Growth Model (Dividend Discount Model): Price of a share = next year's dividend divided by (required return minus dividend growth). A one-line DCF for a company that will grow its dividend at one steady rate forever.
P0 = D1 / (r - g) = D0 x (1+g) / (r - g)Fails when (read this): Non-dividend payers, companies with erratic payout, high-growth firms, and whenever buybacks are the main return channel (understates value). - Graham Number: Graham said a defensive investor should not pay more than 15x earnings and 1.5x book, and that the product of the two multiples should not exceed 22.5. The Graham Number is the price at which that product is exactly 22.5.
Graham Number = sqrt(22.5 x EPS x BVPS)Fails when (read this): Negative EPS or book (returns NaN), asset-light businesses, banks (book value is regulatory), companies with heavy intangibles. - Graham Growth Formula (1962) and 1974 revision: Value = EPS x (8.5 + 2g): a no-growth company deserves 8.5x, and every 1% of expected growth adds 2 points of P/E. The 1974 revision scales this by 4.4/Y, where Y is the AAA corporate bond yield, so higher rates lower value.
V = EPS x (8.5 + 2g) [1962]; V = EPS x (8.5 + 2g) x 4.4 / Y [1974]Fails when (read this): Cyclicals at peak earnings, negative EPS, any g beyond ~15%, and when Y is far from 4.4 (the linear scaling was a US-1970s patch). - Owner Earnings (Buffett): Reported profit + depreciation and other non-cash charges - the capex the business needs just to stay where it is (maintenance capex). It is what an owner could take home without weakening the business.
Owner earnings = Net income + D&A + other non-cash charges - maintenance capex (- increase in working capital)Fails when (read this): Maintenance capex is not disclosed anywhere — every number here is an estimate. Also fails for banks and for companies in a build-out phase. - Residual Income / EVA (Economic Value Added): Profit is only real after you charge for the capital used. EVA = NOPAT - (WACC x capital). Residual income does the same for equity: NI - (k_e x book equity). Value = book value + PV of all future residual income.
EVA = NOPAT - WACC x Invested capital; RI_t = NI_t - k_e x B_{t-1}; V0 = B0 + sum RI_t/(1+k_e)^tFails when (read this): Book value is distorted (goodwill from acquisitions, revaluations, buybacks that shrink equity), tax rates jump, or when capital employed is understated by leases. - Free Cash Flow Yield: Free cash flow divided by market cap (or by EV). It is the 'interest rate' the business pays you if it handed over all its free cash. Compare directly with the 10-year G-Sec.
FCF yield = (OCF - Capex) / Market cap; or FCFF / EVFails when (read this): One-year FCF is lumpy (working-capital swings, capex years); banks; companies capitalising costs to flatter OCF. - Earnings Power Value (EPV): Take normalised, sustainable after-tax operating earnings, assume they never grow, and capitalise them at WACC. EPV is the value of the business as it stands; anything the market pays above EPV is a bet on growth.
EPV = Adjusted EBIT x (1 - t) / WACC (+ excess cash - debt for equity EPV)Fails when (read this): Earnings at cyclical peak or trough, businesses in secular decline (no-growth is optimistic), and where reproduction cost of assets is far from book. - Reverse DCF (Expectations Investing): Run the DCF backwards: hold WACC and terminal growth, and solve for the cash-flow growth rate that makes the DCF equal the current market cap. Then ask one question: is that growth believable?
Find g such that sum FCF0(1+g)^t/(1+r)^t + TV(g)/(1+r)^n = Market value (numerical solve)Fails when (read this): Loss-making or negative-FCF firms (no base to grow), and when r itself is the contested variable rather than g. - Sum-of-the-Parts (SOTP): Value each business segment with the method that suits it (DCF, EV/EBITDA multiple, market value of listed stakes), add them up, subtract net debt and a holding-company discount, and compare with market cap.
Equity = sum(segment EV_i) + listed stakes x (1 - holdco discount) + cash - debt - minority interestFails when (read this): Segments share costs (allocation games), unlisted segment multiples are guesses, and when corporate overhead is ignored (must be capitalised as a negative segment). - Cost of Equity — Bottom-up Beta + Country Risk Premium (Damodaran): Cost of equity = risk-free rate + beta x (mature-market ERP + country risk premium). Beta comes bottom-up (unlever peers' betas, re-lever at your firm's D/E) rather than from a noisy 2-year regression. CRP = India's sovereign default spread scaled by equity/bond volatility.
beta_L = beta_U x [1 + (1-t) x D/E]; CRP = Default spread x (sigma_equity / sigma_bond); k_e = Rf + beta_L x (ERP_mature + CRP)Fails when (read this): Companies with mostly foreign revenue (TCS earns in USD — Damodaran's 'lambda' approach weights CRP by revenue exposure), regression betas on thinly traded small-caps, and whenever the rating changes mid-year. - Piotroski F-Score (9 tests): Nine yes/no accounting tests, one point each, in three groups: profitability (4), leverage/liquidity (3), operating efficiency (2). Piotroski showed that among cheap stocks, high scorers beat low scorers by a wide margin (1976-1996 US sample).
F = [ROA>0] + [CFO>0] + [dROA>0] + [CFO>NI] + [dLeverage<0] + [dCurrentRatio>0] + [no new shares] + [dGrossMargin>0] + [dAssetTurnover>0]Fails when (read this): Designed for cheap (high B/M) stocks — on expensive quality stocks it adds little. Binary tests ignore magnitude; a firm can 'pass' leverage by refinancing short-term. One-year comparisons are noisy for cyclicals. - Altman Z-Score (Z, Z' and Z'' emerging-market): A weighted sum of five (Z, Z') or four (Z'') ratios covering liquidity, retained earnings, operating profitability, leverage and (for the original) asset turnover. Below a cut-off, historical odds of distress within two years were high.
Z = 1.2X1 + 1.4X2 + 3.3X3 + 0.6X4 + 1.0X5 (listed manufacturers; X4 = market equity/total liabilities) Z' = 0.717X1 + 0.847X2 + 3.107X3 + 0.420X4 + 0.998X5 (private firms; X4 = book equity/TL) Z''= 3.25 + 6.56X1 + 3.26X2 + 6.72X3 + 1.05X4 (non-manufacturers & emerging markets; no X5)Fails when (read this): Banks and NBFCs (leverage is the business), very young companies, and companies where the crisis is legal/regulatory rather than financial. It is a 1960s-1990s US/Mexico calibration — treat cut-offs as rough. - Beneish M-Score (8 variables): Eight year-over-year indices (receivables, gross margin, asset quality, sales growth, depreciation, SG&A, leverage, accruals) weighted into one score. Above -1.78 (Beneish's later cut-off; -2.22 in the original) the company looks like the manipulators in his sample.
M = -4.84 + 0.920 DSRI + 0.528 GMI + 0.404 AQI + 0.892 SGI + 0.115 DEPI - 0.172 SGAI + 4.679 TATA - 0.327 LVGIFails when (read this): Financial firms, companies with genuine one-off working-capital jumps (large order wins), firms that changed accounting policy, and any single-year read — trend across 3 years matters more. - Ohlson O-Score: A nine-factor logistic model. Unlike Z-Score it outputs a log-odds that converts to a probability; it adds firm size, a funds-from-operations term and two dummies (negative equity; loss in both of last two years).
O = -1.32 - 0.407 ln(TA/GNP deflator) + 6.03 TL/TA - 1.43 WC/TA + 0.0757 CL/CA - 1.72 OENEG - 2.37 NI/TA - 1.83 FFO/TL + 0.285 INTWO - 0.521 (NI_t - NI_{t-1})/(|NI_t|+|NI_{t-1}|); P(bankruptcy) = e^O / (1 + e^O)Fails when (read this): The GNP deflator term makes SIZE non-comparable across countries and decades; the coefficients are 1970s US. Use it for ranking, not for the probability itself. - Sloan Accrual Ratio: Accruals = profit minus operating cash flow. Sloan showed high-accrual firms see earnings mean-revert and their stocks lag; a long low-accrual / short high-accrual portfolio earned ~12% a year in his sample.
Accrual ratio = (Net income - CFO - CFI_capex-adjusted) / Average total assets; simple form: (NI - CFO) / Avg TAFails when (read this): Fast-growing firms with legitimately rising receivables/inventory, project businesses (percentage-of-completion), banks (CFO meaningless). - Montier C-Score ('Cooking the books'): Six yes/no flags, one point each: (1) net income growing faster than cash flow, (2) rising days sales outstanding, (3) rising days sales of inventory, (4) rising other current assets to revenue, (5) falling depreciation relative to gross PPE, (6) total asset growth above 10%. Montier found high scorers lagged the market by ~8% a year (1993-2007).
C = [NI-CFO gap widening] + [DSO up] + [DSI up] + [Other CA/Sales up] + [Dep/Gross PPE down] + [Asset growth > 10%]Fails when (read this): Acquisitive companies (asset growth flag fires legitimately), businesses moving to longer credit terms by strategy, and seasonal year-ends. - DuPont Analysis — 3-step: ROE = net margin x asset turnover x equity multiplier. Same ROE can come from a pharma company (fat margin), a retailer (fast turnover) or a bank (leverage). The split tells you which — and which is fragile.
ROE = (NI / Sales) x (Sales / Assets) x (Assets / Equity)Fails when (read this): Negative equity (multiplier meaningless), buyback-shrunk equity, and one-off gains in NI. - DuPont Analysis — 5-step: Splits net margin into tax burden x interest burden x operating margin, so ROE = tax burden x interest burden x EBIT margin x asset turnover x equity multiplier. Now you can see whether ROE improved because operations improved or because tax/interest got cheaper.
ROE = (NI/EBT) x (EBT/EBIT) x (EBIT/Sales) x (Sales/Assets) x (Assets/Equity)Fails when (read this): Loss-making firms (ratios flip sign), big non-operating income, and mixed conglomerate segments. - Return on Capital Employed (ROCE): EBIT divided by capital employed (equity + debt, or total assets - current liabilities). It is the pre-tax return the business earns on all long-term money in it, regardless of who funded it.
ROCE = EBIT / (Total assets - Current liabilities)Fails when (read this): Banks/NBFCs (use ROA/ROE), asset-light firms with negative working capital (ROCE explodes), and firms with large cash piles (cash drags ROCE — use ROIC ex-cash). - Return on Invested Capital (ROIC): NOPAT divided by invested capital (equity + debt - excess cash). After-tax, and excluding idle cash, so it measures the return on money actually working in the business. Compare with WACC: the spread is value creation.
ROIC = EBIT x (1 - t) / (Equity + Debt - Excess cash)Fails when (read this): Goodwill treatment changes the answer (include for acquirers' track record, exclude for organic economics), leases, and negative IC (some FMCG/IT with negative working capital). - Cash Return on Invested Capital (CROIC): Free cash flow divided by invested capital. Same denominator as ROIC, but the numerator is cash after capex, so a company that reports high profit but spends it all on plant shows a low CROIC.
CROIC = (CFO - Capex) / Invested capitalFails when (read this): Genuine growth capex years (a new plant is not a treadmill), and lumpy working capital. - Cash Conversion Cycle (CCC): Days inventory + days receivable - days payable. The number of days between paying suppliers and collecting from customers. Shorter is better; negative means suppliers finance the business.
CCC = DIO + DSO - DPO; DIO = Inventory/COGS x 365; DSO = Receivables/Revenue x 365; DPO = Payables/COGS x 365Fails when (read this): Seasonal businesses at year-end (sugar, fertiliser), project accounting, and financials. - Interest Coverage Ratio: EBIT divided by interest expense: how many times operating profit covers the interest bill. Below 1.5x the company is working for the bank; below 1x it is borrowing to pay interest.
ICR = EBIT / Interest expense (or EBITDA / Interest for capex-heavy firms)Fails when (read this): Capitalised interest (construction phase) hides the true bill; zero-debt firms (ratio infinite, meaningless); financials. - Promoter Pledge % (India-specific): Share of the promoter's own holding that is pledged (or otherwise encumbered) as loan collateral, and the same as a share of total company equity. Falling prices trigger margin calls, lenders sell, price falls more — a reflexive loop.
Pledge% (of promoter holding) = Pledged shares / Promoter shares; Pledge% (of company) = Pledged shares / Total sharesFails when (read this): Disclosure is event-based and can lag; NDU (non-disposal undertakings) and unlisted holdco-level pledges may not show. Zero pledge does not mean zero promoter leverage. - Related-Party Transaction (RPT) Ratio (India-specific): Total related-party transactions (sales, purchases, loans, guarantees, royalties) divided by consolidated revenue. Measures how much of the business runs through entities the promoter controls, where pricing is not at arm's length by default.
RPT ratio = Sum of RPT value (Ind AS 24 note) / Consolidated revenue; also track RPT loans+guarantees / Net worthFails when (read this): Group companies with genuine vertical integration (auto ancillaries supplying the parent), and JVs; also captive-user PSUs. Judge purpose, not just size. - CAN SLIM (7 components): Seven traits O'Neil found in past big winners: C - current quarterly EPS up 25%+ YoY; A - annual EPS growth 25%+ over 3 years; N - something new (product, management, new high); S - supply/demand (volume on up days, tight float); L - leader in its group (relative strength); I - institutional sponsorship rising; M - market direction up.
Score = [C: qEPS growth >= 25%] + [A: 3-yr EPS CAGR >= 25%] + [N: new high / new catalyst] + [S: volume up on advances] + [L: RS rank >= 80] + [I: # institutions rising] + [M: index above 50-DMA & uptrend]Fails when (read this): Cyclicals at turning points (earnings growth lags price), companies with lumpy order books, and bear markets (M rule keeps you out — by design). - Minervini Trend Template (8 criteria) + SEPA: Eight price-structure tests that must ALL pass: (1) price > 150-DMA and > 200-DMA; (2) 150-DMA > 200-DMA; (3) 200-DMA rising for at least 1 month (ideally 4-5); (4) 50-DMA > 150-DMA > 200-DMA; (5) price > 50-DMA; (6) price at least 25-30% above 52-week low; (7) price within 25% of 52-week high; (8) RS rank >= 70 (leaders 90+). SEPA is his wider method: Trend, Fundamentals, Catalyst, Entry point, Exit point.
Pass = P>MA150 & P>MA200 & MA150>MA200 & MA200 rising(>=1m) & MA50>MA150>MA200 & P>MA50 & P>=1.25*Low52 & P>=0.75*High52 & RS>=70Fails when (read this): Whipsaw markets (2022-type), stocks with < 1 year of listing (no 200-DMA or 52-week data), and illiquid small-caps where MAs are noisy. - Darvas Box: A stock making 52-week highs trades in a 'box' between a recent high and low. Buy when price breaks above the box top on rising volume; place the stop just under the box bottom; when a new box forms higher, raise the stop. Never buy inside the box, never average down.
Box top = highest high not exceeded for 3 days; Box bottom = lowest low after the top not broken for 3 days; Buy if Close > Box top (volume > avg); Stop = Box bottom - small bufferFails when (read this): Sideways markets (repeated false breakouts), stocks with circuit limits (box edges become artificial), and gap-heavy stocks where stops are jumped. - Lynch's Six Categories + PEG Rule: First sort the stock into one of six boxes — slow grower, stalwart, fast grower (20%+ EPS growth), cyclical, turnaround, asset play — because each is analysed differently. For growers, use PEG = P/E divided by EPS growth rate: near 1 is fair, below 1 attractive, above 2 expensive. Lynch's dividend-adjusted version adds yield to growth.
PEG = (P/E) / (EPS growth % ); Lynch dividend-adjusted: (P/E) / (EPS growth % + Dividend yield %)Fails when (read this): Cyclicals (PEG is meaningless at peak margins), turnarounds and asset plays (no g), negative or tiny EPS, and growth above ~40% (PEG breaks down). - Coffee Can Screen (Marcellus filter): Marcellus's filter: revenue growth of at least 10% AND pre-tax ROCE of at least 15% in EVERY one of the last 10 years (not on average). The point is consistency — a single miss disqualifies. The universe that passes is small (Marcellus reports ~25-30 companies).
Pass = AND over t=1..10 of [RevGrowth_t >= 10%] AND [ROCE_t >= 15%] (financials: loan growth >= 15% and ROE >= 15% in Marcellus's variant)Fails when (read this): Cyclicals and commodity businesses by construction; young companies (< 10 years); and it says nothing about price — a Coffee Can stock at 80x earnings is still 80x. - 100-Baggers Twin Engine (Phelps / Mayer): Return multiple = EPS growth multiple x P/E expansion multiple. A 100x needs both engines: e.g. earnings 20x and multiple 5x. Mayer's ingredients: high ROIC (20%+) reinvested for a long time, small starting size, owner-operators, and a holding period of 15-25 years.
Price multiple = (EPS_end / EPS_start) x (PE_end / PE_start); Years to 100x at CAGR r = ln(100)/ln(1+r)Fails when (read this): Survivorship: Mayer's 365 are the winners; the base rate is a fraction of a percent of listed stocks (see Bessembinder). It is a lens for holding, not a screen for finding. - Bessembinder: Do Stocks Outperform Treasury Bills? (return skew): Using all ~25,300 US common stocks in CRSP from 1926 to 2016, Bessembinder found that only 42.6% of stocks had a lifetime buy-and-hold return above one-month Treasury bills; the median stock lost money; the best-performing 4% of companies account for the entire net wealth creation of the market (~$35 trillion as of Dec 2016) because the other 96% collectively matched T-bills; and just 86 firms produced half of that wealth. The market's positive average return comes from a tiny right tail, not from the typical stock.
Wealth creation_i = Terminal value of $1 in stock i - Terminal value of $1 in T-bills, summed over its listed life; Share of positive-sum firms needed to explain total = 4%Fails when (read this): It is a US 1926-2016 result; a 2019 follow-up on 42 countries found similar or stronger skew abroad. It does not say active management cannot work — it says the base rate is against the typical pick. - Markowitz mean-variance portfolio: Finds the mix of assets that gives the lowest variance for a chosen expected return. Diversification works because assets do not move together perfectly - the covariance term is the whole point.
E[Rp] = sum(w_i * mu_i); Var(Rp) = sum_i sum_j w_i w_j sigma_ij; minimise Var(Rp) subject to sum(w_i)=1 and E[Rp]=targetFails when (read this): Inputs are estimates: expected returns are very noisy, and the optimiser leans hardest on the noisiest numbers ('error maximiser'). Covariances jump in crashes exactly when you need diversification. - Efficient frontier and tangency (max-Sharpe) portfolio: The set of portfolios with the highest return for each level of risk. Add a risk-free asset and the best risky portfolio is the one where a line from the risk-free rate touches the frontier - the maximum-Sharpe (tangency) portfolio.
Tangency weights: w* = Sigma^-1 (mu - rf*1) / (1' Sigma^-1 (mu - rf*1)); Sharpe = (E[Rp]-rf)/sigma_p; CML: E[R] = rf + sigma * Sharpe_tangencyFails when (read this): Tangency weights explode with tiny changes in expected returns; without constraints they often demand huge shorts. Sample covariance with more assets than years of data is singular. - CAPM - Capital Asset Pricing Model: Says the only risk you get paid for is market risk (beta). Expected return = risk-free + beta x market premium. Everything else is diversifiable and earns nothing.
E[Ri] = rf + beta_i * (E[Rm] - rf); beta_i = Cov(Ri,Rm)/Var(Rm)Fails when (read this): Empirically the security market line is too flat: low-beta stocks earn more than CAPM says and high-beta less (Black-Jensen-Scholes 1972; Frazzini-Pedersen 2014). Beta is unstable across windows and regimes; single-factor cannot explain size, value, momentum. - Fama-French three-factor model: Adds two factors to CAPM: SMB (small minus big) and HML (high minus low book-to-market). A fund's 'alpha' is what is left after these three are paid for.
Ri - rf = alpha + b*(Rm - rf) + s*SMB + h*HML + eFails when (read this): Factors are empirical, not theory; the 1993 paper itself says why they are priced is unclear. Momentum is missing (hence Carhart). Loadings drift, and short windows give meaningless alphas with wide error bars. - Fama-French five-factor model: Adds profitability (RMW: robust minus weak) and investment (CMA: conservative minus aggressive) to the three-factor model. In the 2015 tests HML becomes largely redundant once RMW and CMA are in.
Ri - rf = alpha + b*(Rm-rf) + s*SMB + h*HML + r*RMW + c*CMA + eFails when (read this): Still fails to price small stocks that invest a lot despite low profitability (the paper's own admission). Momentum still absent. Six-factor and q-factor (Hou-Xue-Zhang 2015) alternatives compete. - Carhart four-factor (momentum, UMD/WML): Adds a one-year momentum factor (winners minus losers, WML or UMD) to Fama-French three. Carhart showed most 'hot hand' fund persistence was just momentum exposure plus expenses.
Ri - rf = alpha + b*(Rm-rf) + s*SMB + h*HML + m*WML + e; WML = avg return of top 30% past 12-2 month winners minus bottom 30% losersFails when (read this): Momentum crashes (2009 in US, Apr-Jun 2009 and 2020 in India) hit WML with -30% to -50% months; the factor is priced but fat-tailed. Skip-month and formation windows change results. - Cross-sectional momentum (12-1 formation): Rank stocks by past 12-month return excluding the most recent month; past winners tend to keep winning for 3-12 months. The skipped month avoids short-term reversal.
MOM_12_1 = P(t-1) / P(t-12) - 1; rank cross-section; long top decile, short bottom decile; hold 1-3 monthsFails when (read this): Momentum crashes after sharp bear-market rebounds when the short leg (beaten-down high-beta names) rallies violently. High turnover means costs matter more than in any other factor. - Betting Against Beta (BAB): Because many investors cannot borrow, they overpay for high-beta stocks to get exposure. BAB goes long low-beta assets levered up to beta 1 and short high-beta assets de-levered to beta 1; the spread has been positive across 20 countries.
r_BAB = (1/beta_L) * (r_L - rf) - (1/beta_H) * (r_H - rf); beta_i = rho_im * sigma_i / sigma_m (shrunk: 0.6*beta_TS + 0.4*1)Fails when (read this): Needs cheap leverage and the ability to short; when funding is tight (2008, Mar 2020) BAB draws down as leveraged low-beta holders de-lever. Beta estimates are noisy for thin stocks. - Quality Minus Junk (QMJ): Scores stocks on profitability, growth and safety (z-scored and averaged), then goes long the top 30% and short the bottom 30%. Quality is priced too cheaply - high-quality firms earn more than their higher valuation implies.
Quality = z( z(Profitability) + z(Growth) + z(Safety) ); Profitability = z(GPOA)+z(ROE)+z(ROA)+z(CFOA)+z(GMAR)+z(ACC); QMJ = 0.5*(SmallQuality + BigQuality) - 0.5*(SmallJunk + BigJunk)Fails when (read this): Quality is a defensive factor: it lags in sharp junk rallies (2009, late 2020-21 in India). Accounting inputs are backward-looking and manipulable. - Black-Litterman model (structure): Starts from the returns the market itself implies (reverse-optimised from cap weights), then tilts them toward your views in proportion to how confident you are. Fixes Markowitz's wild weights by anchoring to equilibrium.
Pi = lambda * Sigma * w_mkt; E[R] = [ (tau Sigma)^-1 + P' Omega^-1 P ]^-1 [ (tau Sigma)^-1 Pi + P' Omega^-1 Q ]Fails when (read this): Equilibrium depends on choosing the right market portfolio and lambda; tau and Omega are conventions with no agreed values, and results are sensitive to them. Views are still forecasts. - Risk parity / inverse-volatility weighting: Sizes positions so each asset contributes the same risk, instead of the same rupees. A 60/40 equity/bond portfolio has about 90% of its risk in equity; risk parity fixes that by holding more of the calm asset.
Naive (zero-correlation) version: w_i = (1/sigma_i) / sum_j (1/sigma_j). Full version solves w_i * (Sigma w)_i = equal for all i.Fails when (read this): Assumes the low-vol asset keeps diversifying; in 2022 bonds and equities fell together and levered risk parity suffered. Vol estimates lag regime changes. - Kelly-optimal fraction (continuous / portfolio form): The fraction of capital that maximises long-run growth rate. For a continuous asset it is excess return over variance; for many assets it is Sigma^-1 times excess returns - the same direction as the tangency portfolio, but with an absolute size.
f* = (mu - rf) / sigma^2 (single asset); f_vec = Sigma^-1 (mu - rf 1) (portfolio); growth g = rf + f(mu-rf) - 0.5 f^2 sigma^2; half-Kelly = f*/2Fails when (read this): Assumes you know mu and sigma and that returns are continuous with no gaps; real markets gap, and overestimating mu by 2x turns Kelly into ruin. Ignores your own utility (most people cannot stomach Kelly drawdowns). - Factor crowding caveat (valuation spread and decay): Not a return formula - a warning with numbers. Published anomalies lose about a third to a half of their return after publication (McLean-Pontiff: 26% lower out of sample, 58% lower post-publication). A factor's valuation spread (long-leg vs short-leg multiples) tells you how crowded it is.
Post-pub return ~ (1 - 0.58) * in-sample return; Valuation spread = log( P/B_short_leg / P/B_long_leg ); compare spread to its own history (z-score)Fails when (read this): Decay estimates are US averages; individual factors differ. Valuation spread is a slow, weak timing signal that can stay extreme for years. - Simple Moving Average (SMA) and crossovers: Average of the last n closes; smooths noise so the trend shows. Golden cross = shorter SMA rising above longer (classic 50 over 200); death cross = the reverse.
SMA_n = (C1 + C2 + ... + Cn) / n; crossover when SMA_short crosses SMA_longFails when (read this): Lags by about n/2 bars and whipsaws in sideways markets; every crossover system loses money in ranges and only pays in the few long trends. Equal weighting means a big old bar dropping out moves the average even if today is quiet. - Exponential Moving Average (EMA): A moving average that weights recent closes more, with weights decaying geometrically. Reacts faster than the SMA of the same length and never fully forgets old data.
k = 2/(n+1); EMA_t = k*C_t + (1-k)*EMA_(t-1); seed EMA with SMA of first n closesFails when (read this): Different platforms seed differently (SMA seed vs first close), so EMAs differ for months on short histories. Faster response also means more false crossovers in chop. - Weighted Moving Average (WMA): Linear weights: the newest close gets weight n, the oldest gets 1. Between SMA and EMA in responsiveness, and unlike EMA it fully drops data older than n.
WMA_n = (n*C_t + (n-1)*C_(t-1) + ... + 1*C_(t-n+1)) / (n(n+1)/2)Fails when (read this): Same whipsaw problem as any average; the linear weighting is arbitrary, not derived from anything. - MACD (Moving Average Convergence/Divergence): MACD line = 12-EMA minus 26-EMA (the distance between two averages). Signal = 9-EMA of MACD. Histogram = MACD minus signal. Rising histogram means the short average is pulling away from the long one - momentum.
MACD = EMA_12(C) - EMA_26(C); Signal = EMA_9(MACD); Histogram = MACD - SignalFails when (read this): It is a difference of two lagging averages - in a range it flips constantly. Absolute value is price-scaled, so you cannot compare MACD across stocks (use PPO for that). Divergences can persist for many bars. - RSI - Relative Strength Index (Wilder smoothing): Ratio of average up-closes to average down-closes over 14 periods, squashed to 0-100. Wilder's smoothing is a specific recursion (previous average x 13 + today) / 14 - it is NOT a simple average after the first value, and NOT the 2/(n+1) EMA.
U = max(C_t - C_(t-1), 0); D = max(C_(t-1) - C_t, 0); first AvgU = mean(U over n); AvgU_t = (AvgU_(t-1)*(n-1) + U_t)/n (same for D); RS = AvgU/AvgD; RSI = 100 - 100/(1+RS)Fails when (read this): Overbought can stay overbought for weeks in a trend; RSI is a range tool that fails as a reversal signal in trends. Values depend on the seed and history length - two platforms with different history give different RSI for the same day. - Stochastic oscillator (%K, %D) - Lane: Where today's close sits inside the high-low range of the last n bars, 0 to 100. Lane's idea: in uptrends closes cluster near the highs, in downtrends near the lows, and momentum turns before price.
%K = 100 * (C - L_n) / (H_n - L_n); %D = SMA_3(%K); 'slow' stochastic uses SMA_3(%K) as %K and SMA_3 of that as %DFails when (read this): Pinned at 80+ or 20- through entire trends; the divergence signal Lane actually traded needs patience most users lack. Range denominators shrink to near zero in quiet weeks, making %K jumpy. - Bollinger Bands, %B and Bandwidth: A 20-period SMA with bands 2 standard deviations above and below. %B says where price is between the bands; Bandwidth measures how wide they are. Bollinger's rule: the bands are relative high and low, not signals by themselves.
Mid = SMA_20; Upper = Mid + 2*sigma_20; Lower = Mid - 2*sigma_20 (population sd); %B = (C - Lower)/(Upper - Lower); Bandwidth = (Upper - Lower)/MidFails when (read this): Touching a band is not a reversal signal - in trends price 'walks the band'. Two sigma covers ~95% only for normal data; returns are fat-tailed so breaches are more common than the number suggests. - ATR - Average True Range (Wilder smoothing): True Range is the largest of: today's high-low, |high - previous close|, |low - previous close| - so gaps count. ATR is Wilder's smoothed average of TR over 14 periods. It measures how far the instrument typically moves, in price units.
TR = max(H - L, |H - C_prev|, |L - C_prev|); ATR_t = (ATR_(t-1)*(n-1) + TR_t)/n, seeded with mean TR over first nFails when (read this): Absolute units - ATR 300 on NIFTY and ATR 3 on a small cap are not comparable (divide by price). Lags after regime shifts because of the 14-period memory. Says nothing about direction. - ADX / DMI (+DI, -DI) - Wilder: +DM is today's high minus yesterday's high (if positive and larger than the down move); -DM the mirror. Divide the smoothed DMs by ATR to get +DI and -DI. DX measures their separation; ADX is the Wilder-smoothed DX. ADX tells you trend STRENGTH, not direction - the DIs give direction.
+DM = H_t - H_(t-1) if > (L_(t-1) - L_t) and > 0 else 0; -DM = L_(t-1) - L_t if > (H_t - H_(t-1)) and > 0 else 0; +DI = 100 * Wilder(+DM,14)/ATR_14; -DI likewise; DX = 100*|+DI - -DI|/(+DI + -DI); ADX = Wilder(DX,14)Fails when (read this): Very laggy (two layers of 14-period smoothing); ADX often peaks after the trend has ended. DI crossovers alone are a poor entry signal. - Supertrend: An ATR-based trailing line: basic upper band = (H+L)/2 + m*ATR, lower = (H+L)/2 - m*ATR. The line ratchets - the lower band only rises in an uptrend, the upper band only falls in a downtrend - and flips when close crosses it. Defaults on most Indian platforms: period 10, multiplier 3.
BU = (H+L)/2 + m*ATR_n; BL = (H+L)/2 - m*ATR_n; FinalLower_t = max(BL_t, FinalLower_(t-1)) if C_(t-1) > FinalLower_(t-1) else BL_t; FinalUpper_t = min(BU_t, FinalUpper_(t-1)) if C_(t-1) < FinalUpper_(t-1) else BU_t; Supertrend = FinalLower in uptrend, FinalUpper in downtrend; flip when close crossesFails when (read this): Whipsaws badly in ranges - many small losses. Not a published, peer-reviewed method; parameters (10,3) are folklore. Implementations differ on whether they use close or high/low to flip. - Donchian channels (n-bar breakout): Upper line = highest high of the last n bars, lower = lowest low, middle = their average. Buy a break above the upper line, exit on a break of a shorter lower line. The Turtles used 20/10 and 55/20.
Upper_n = max(H over n); Lower_n = min(L over n); Mid = (Upper+Lower)/2Fails when (read this): Win rate ~30-40%; long strings of false breakouts in ranges. Needs discipline to take every signal - skipping one often skips the trend. - Ichimoku Kinko Hyo (five lines): Five lines built from midpoints of ranges, with fixed periods 9, 26, 52 (Hosoda's numbers from the six-day Japanese trading week). Two lines are shifted forward 26 bars to form the cloud; one (the lagging span) is today's close plotted 26 bars back.
Tenkan-sen = (max H_9 + min L_9)/2; Kijun-sen = (max H_26 + min L_26)/2; Senkou Span A = (Tenkan + Kijun)/2 plotted 26 ahead; Senkou Span B = (max H_52 + min L_52)/2 plotted 26 ahead; Chikou Span = close plotted 26 backFails when (read this): Needs 52+26 = 78 bars of history to draw fully; useless on new listings. Thin, choppy cloud gives constant false breaks. The 9/26/52 periods were tuned to a 6-day week and are kept by tradition, not testing. - VWAP and Anchored VWAP: Cumulative sum of price x volume divided by cumulative volume, reset each session (09:15 on NSE). It is the average price everyone paid so far today. Anchored VWAP starts the sum at a chosen bar - an earnings day, a swing low - instead of the open.
VWAP_t = sum_(i<=t) (TP_i * V_i) / sum_(i<=t) V_i, TP = (H+L+C)/3, reset at session open; AVWAP: same sum starting at anchor barFails when (read this): Meaningless on daily charts without a reset rule; dominated by the opening 15 minutes on NSE where a large share of the day's volume prints. AVWAP depends entirely on the anchor choice. - Pivot points - Classic, Fibonacci, Camarilla: Turns yesterday's high, low and close into today's reference levels. Classic uses the pivot P = (H+L+C)/3 and reflects the range around it; Fibonacci scales the range by 0.382/0.618/1.0; Camarilla hugs the close with 1.1 x range / 12, 6, 4, 2.
Classic: P=(H+L+C)/3; R1=2P-L; S1=2P-H; R2=P+(H-L); S2=P-(H-L). Fibonacci: R1=P+0.382(H-L); R2=P+0.618(H-L); R3=P+(H-L); S mirror. Camarilla: R3=C+1.1(H-L)/4; R4=C+1.1(H-L)/2; S3=C-1.1(H-L)/4; S4=C-1.1(H-L)/2Fails when (read this): Self-fulfilling only to the extent others watch the same formula; no statistical edge on its own. After a gap open all levels are stale. Different sites use different variants (some include open in P). - Fibonacci retracement and extension: Marks how much of a swing has been given back: 23.6%, 38.2%, 50%, 61.8%, 78.6%. Extensions (127.2%, 161.8%) project beyond the swing. The ratios come from limits of consecutive Fibonacci numbers (F(n)/F(n+1) -> 0.618, F(n)/F(n+2) -> 0.382); 50% and 78.6% (= sqrt 0.618) are added by convention.
Up-swing from L to H: retracement level = H - r*(H-L), r in {0.236,0.382,0.5,0.618,0.786}; extension = H + e*(H-L), e in {0.272,0.618,1.0}; phi = (1+sqrt5)/2 = 1.618Fails when (read this): Swing selection is subjective - two traders draw different levels. No peer-reviewed evidence that prices respect 61.8% more than any other level; with five levels drawn, price 'hits' one of them by chance. - Elliott Wave - the three inviolable rules: Trends unfold in five waves (1-2-3-4-5) against three corrective waves (a-b-c). Only three rules are absolute; everything else is a guideline. Rule 1: wave 2 never retraces more than 100% of wave 1. Rule 2: wave 3 is never the SHORTEST of waves 1, 3 and 5 (it is often the longest, but 'must be longest' is wrong). Rule 3: in an impulse, wave 4 does not enter the price territory of wave 1 (diagonals are the exception).
Rule 1: |W2| < |W1| (end of 2 above start of 1 in an uptrend). Rule 2: len(W3) > min(len(W1), len(W5)). Rule 3: low of W4 > high of W1 (impulse, uptrend). Guidelines only: W2 ~ 0.5-0.618 of W1; W4 ~ 0.382 of W3; W3 ~ 1.618 x W1; alternation between 2 and 4.Fails when (read this): Subjective wave identification; counts get revised after the fact, so the method is nearly unfalsifiable in real time. Degree labelling (minute vs minor) is arbitrary. No peer-reviewed out-of-sample evidence. - Dow Theory - six tenets: The root of Western technical analysis. Six tenets: (1) the averages discount everything; (2) the market has three trends - primary (years), secondary (weeks-months), minor (days); (3) primary trends have three phases - accumulation, public participation, distribution; (4) the averages must confirm each other (Dow used Industrials and Rails); (5) volume must confirm the trend; (6) a trend is assumed to continue until definite reversal signals appear.
Uptrend: successive higher highs AND higher lows on the primary index, confirmed by a second index and by volume expanding in the trend direction. Reversal: a lower low after a failed higher high (or vice versa) on both averages.Fails when (read this): Signals arrive late - Dow theory typically confirms a reversal 20-25% into the new trend. Which averages count and what 'definite reversal' means are judgement calls; Rhea and Hamilton disagreed on specifics. - Wyckoff accumulation / distribution and the Composite Man: Read the market as if one 'Composite Man' were running it: he accumulates quietly in a range, marks price up, distributes to the public at the top, marks down. Accumulation schematic events: PS (preliminary support), SC (selling climax), AR (automatic rally), ST (secondary test), Spring (a false break below the range), SOS (sign of strength), LPS (last point of support). Distribution mirrors: PSY, BC (buying climax), AR, ST, UT/UTAD (upthrust), SOW, LPSY. Three laws: supply and demand, cause and effect (range width -> move size), effort vs result (volume vs price progress).
Effort vs result: if volume (effort) is high but the bar's net progress (result) is small, the prevailing move is being absorbed. Spring = low < range low, then close back inside on falling volume. Cause-and-effect count (P&F): projected move = range width x box x reversal.Fails when (read this): Labels are assigned in hindsight; the same range can be read as accumulation or distribution until it resolves. Volume interpretation is qualitative. The 'Composite Man' is a teaching metaphor, not a testable claim. - Market Profile / Volume Profile - TPO, POC, Value Area: Stacks the day's price action by price instead of by time: each 30-minute period is a letter (TPO); the price with the most TPOs (or most volume, in Volume Profile) is the Point of Control (POC); the Value Area is the narrowest band around the POC holding 70% of TPOs (about one standard deviation). Shapes - normal, trend, double-distribution - describe the auction.
TPO count per price level; POC = argmax(count); Value Area: start at POC, add the adjacent level (above or below) with the larger count, repeat until cumulative count >= 0.70 * total; VAH/VAL = top/bottom of that bandFails when (read this): Level width (tick size grouping) and the 70% cut-off change the VA materially. TPO vs volume profile disagree on thin instruments. It describes what happened; the trade rules built on it are the practitioner's, not Steidlmayer's. - Mansfield Relative Strength and sector-relative RS: Relative strength line = stock price / index price. Mansfield RS normalises that ratio against its own 52-week moving average so the number is zero-centred: positive means the stock has been beating the index over the past year, negative means lagging. Sector-relative RS uses the sector index as the denominator instead.
RS_t = P_stock,t / P_index,t; Mansfield RS = ( RS_t / SMA_52w(RS) - 1 ) * 10 (Weinstein plots weekly; some charts omit the x10)Fails when (read this): Relative strength can be positive while the stock is falling (it just falls less than the index). Sector-relative RS hides sector-wide collapses. The 52-week window is slow to flag new leadership. - Heikin-Ashi candle construction: Averaged candles: the close is the average of O,H,L,C; the open is the midpoint of the previous HA candle; high and low are the extremes of the real high/low and the HA open/close. Trends show as runs of same-colour candles with no lower wicks (up) or no upper wicks (down).
HA_Close = (O+H+L+C)/4; HA_Open = (HA_Open_prev + HA_Close_prev)/2 (first bar: (O+C)/2); HA_High = max(H, HA_Open, HA_Close); HA_Low = min(L, HA_Open, HA_Close)Fails when (read this): HA values are not tradable prices - the HA close is not where you can exit. Because the open is recursive, HA lags and the first bars depend on where you start the series. Stops must be set on real prices. - Renko brick construction: Ignores time. A new brick of fixed size prints only when price moves a full brick beyond the last brick; a reversal brick needs a move of two brick sizes from the last brick's edge. Bricks are usually built from closes; ATR-sized bricks are a common modern variant.
Brick size b. Up brick when C >= top_last + b (draw from top_last to top_last+b, repeat while true). Reversal down when C <= top_last - 2b (draw brick from top_last - b to top_last - 2b). Mirror for downtrend. Time axis is bricks, not days.Fails when (read this): Brick size is everything: too small reproduces the noise, too large misses the move. Intrabar movement is invisible, so a stop based on bricks can be hit far past the brick. Backtests on Renko often look at the brick close, which is unattainable in practice. - OBV - On-Balance Volume: A running total of volume: add the whole day's volume on an up-close, subtract it on a down-close, unchanged on a flat close. The absolute number means nothing; its direction versus price is the message - Granville's 'volume precedes price'.
OBV_t = OBV_(t-1) + V_t if C_t > C_(t-1); OBV_(t-1) - V_t if C_t < C_(t-1); OBV_(t-1) if equalFails when (read this): A 0.01% up-close counts the entire day's volume as buying - crude. One block deal or index rebalancing day dominates the line for weeks. Divergences can persist through whole trends. - MFI - Money Flow Index: RSI built on money flow instead of price change. Money flow = typical price x volume; classify each day positive or negative by whether typical price rose; ratio the 14-day sums; squash to 0-100. Unlike RSI it uses simple sums, not Wilder smoothing.
TP = (H+L+C)/3; MF = TP*V; PosMF = sum of MF on days TP > TP_prev over n; NegMF = sum on days TP < TP_prev; MR = PosMF/NegMF; MFI = 100 - 100/(1+MR)Fails when (read this): Same trend-pinning as RSI. Sensitive to single huge-volume days (block deals). The whole day's money flow is signed by a tiny TP change. - CMO - Chande Momentum Oscillator: Sum of up-moves minus sum of down-moves, divided by their total, over n bars, scaled to -100..+100. Like RSI without the smoothing and centred on zero - so it reaches extremes faster and shows both directions symmetrically.
CMO_n = 100 * (Su - Sd) / (Su + Sd), where Su = sum of up-closes over n, Sd = sum of |down-closes| over nFails when (read this): No smoothing means it is noisy; a single large bar dominates a short window (the 269.95 drop here is 68% of S_d). Same trend-pinning issue as RSI. - KAMA - Kaufman Adaptive Moving Average: An EMA whose speed changes with the Efficiency Ratio: net price change over n bars divided by the sum of absolute bar-to-bar changes. ER near 1 = clean trend, so KAMA moves fast (like a 2-period EMA); ER near 0 = noise, so it barely moves (like a 30-period EMA).
ER = |C_t - C_(t-n)| / sum_(i=1..n) |C_(t-i+1) - C_(t-i)|; SC = [ER*(2/(2+1) - 2/(30+1)) + 2/(30+1)]^2; KAMA_t = KAMA_(t-1) + SC*(C_t - KAMA_(t-1))Fails when (read this): At the start of a trend ER is still low, so KAMA is late; at the end of a trend ER is high, so it follows price down before flattening. The 2/30 speed bounds are Kaufman's choices, not derived. - ROC - Rate of Change / Momentum: Percent change over n bars. The simplest momentum line; the 12-month version is the raw material of the academic momentum factor.
ROC_n = (C_t / C_(t-n) - 1) * 100; Momentum_n = C_t - C_(t-n)Fails when (read this): Drop-off effect: ROC jumps when a big old bar leaves the window even if today is quiet. Unbounded, so no fixed overbought level. - Keltner channel (Chester Keltner; ATR version by Linda Raschke): An EMA with bands a fixed number of ATRs above and below (modern form: 20-EMA, 2 x ATR(10)). Because ATR changes slowly, the channel is smoother than Bollinger; Bollinger bands inside the Keltner channel is the 'squeeze' setup.
Mid = EMA_20(C); Upper = Mid + k*ATR_10; Lower = Mid - k*ATR_10 (k = 2 typical)Fails when (read this): Squeeze releases fail as often as they run; the direction must come from elsewhere. Parameters are folklore. - Correlation vs Cointegration: Correlation measures whether two return series move together day to day. Cointegration asks a different question: does a fixed combination of the two PRICE levels stay stationary (mean-reverting) over time? Pairs trading needs the second, not the first.
rho = Cov(rA, rB) / (sigmaA * sigmaB) [returns]; cointegrated if e_t = P_A - beta*P_B is I(0) while P_A, P_B are I(1)Fails when (read this): Structural breaks (merger, index exit, regulatory shock) kill cointegration silently; both measures are backward-looking and sample-dependent. - Engle-Granger Two-Step Cointegration Test: Step 1: regress one price on the other to get the hedge ratio. Step 2: test the residual for a unit root. If the residual is stationary the pair is cointegrated. Simple, but the choice of which stock goes on the left changes the answer.
Step1: P_A = alpha + beta*P_B + e_t (OLS). Step2: ADF on e_t with Engle-Granger critical values (not standard ADF tables). Error correction: dP_A = gamma*e_{t-1} + ...Fails when (read this): Uses standard ADF tables by mistake (too lenient); more than two assets (use Johansen); parameter drift; small samples (< 250 obs) give unreliable betas. - Johansen Cointegration Test (structure): A VAR-based test that finds how many independent stationary combinations exist among N price series, and gives the weights. Unlike Engle-Granger it does not depend on which series is 'dependent', and it handles baskets of 3+.
dY_t = Pi*Y_{t-1} + sum Gamma_i dY_{t-i} + e_t ; rank(Pi)=r = number of cointegrating vectors; Pi = alpha*beta'. Trace stat = -T * sum_{i=r+1..N} ln(1-lambda_i); Max-eigen stat = -T ln(1-lambda_{r+1})Fails when (read this): Sensitive to lag length and deterministic-term choice; over-rejects in small samples; assumes Gaussian errors; weights from beta can be economically absurd (huge short in an illiquid name). - Augmented Dickey-Fuller Test: Tests whether a series has a unit root (random walk) or mean-reverts. The null is 'unit root'; a very negative t-statistic rejects it. The 'augmented' lags soak up autocorrelation.
dy_t = a + b*t + gamma*y_{t-1} + sum_{i=1..p} delta_i*dy_{t-i} + e_t ; H0: gamma = 0 ; test stat = gamma_hat / SE(gamma_hat) compared with Dickey-Fuller tablesFails when (read this): Low power against near-unit-root processes; wrong lag count; structural breaks look like unit roots; using it on the residual of an OLS regression needs Engle-Granger critical values instead. - Z-Score Entry / Exit Rule for Spreads: Standardises today's spread against its rolling mean and standard deviation. Enter when the spread is stretched (|z| above a threshold), exit when it returns to zero, stop if it stretches further than a cut-off.
z_t = (S_t - mean_N(S)) / sd_N(S). Enter short spread if z > +2, long spread if z < -2; exit at |z| < 0.5 (or 0); stop at |z| > 3.5 or after 2x half-life.Fails when (read this): Rolling mean and sd lag a true shift in the spread's level, so the rule keeps adding to a broken pair; thresholds tuned in-sample overfit. - Ornstein-Uhlenbeck Process & Half-Life: Models a spread that is pulled back to a long-run mean at speed theta while being shaken by noise. Half-life = ln2/theta tells you how many days it takes for half of a deviation to decay on average.
dS = theta*(mu - S)dt + sigma dW. Discrete: S_t - S_{t-1} = a + b*S_{t-1} + e; theta = -b (per period); half-life = ln(2)/theta.Fails when (read this): theta estimate is biased upward in small samples (looks faster than it is); assumes constant mu — a drifting mean gives a fake half-life. - Hurst Exponent: Measures whether a series trends (H>0.5), random-walks (H=0.5) or mean-reverts (H<0.5) by looking at how its range grows with the time window. A hydrologist built it for Nile floods; traders use it to decide which playbook a spread deserves.
E[R(n)/S(n)] = C * n^H -> H = slope of log(R/S) vs log(n). Variance-ratio form: Var(y_{t+tau}-y_t) ~ tau^{2H}.Fails when (read this): Very sample-dependent; short-window estimates are biased toward 0.5; H is a description of the past, not a forecast; different estimators (R/S, DFA, variance ratio) disagree. - Kalman Filter Dynamic Hedge Ratio: Treats the hedge ratio as a hidden state that drifts slowly, and updates it every day from the new price pair with a weight (the Kalman gain) that balances trust in the model against trust in the new observation. Replaces rolling OLS with something that reacts faster and forgets smoothly.
State: beta_t = beta_{t-1} + w_t (Q). Obs: y_t = beta_t * x_t + v_t (R). Predict: beta_pred = beta_{t-1}, P_pred = P_{t-1} + Q. Gain K = P_pred*x_t / (x_t^2 P_pred + R). Update: beta_t = beta_pred + K*(y_t - beta_pred*x_t); P_t = (1 - K x_t) P_pred.Fails when (read this): Q and R are hand-tuned and effectively decide the strategy; a linear-Gaussian model cannot capture a sudden break, it only catches up to it. - Distance Method vs Copula Method: Distance method: pick pairs whose normalised price paths have the smallest sum of squared differences, trade at 2 sigma. Copula method: model the joint distribution of the two returns with a copula and trade when one asset's conditional probability given the other is extreme. Copulas capture asymmetric tail co-movement that the distance rule ignores.
Distance: SSD = sum_t (Pn_A,t - Pn_B,t)^2 with Pn normalised to 1 at start. Copula: h1 = P(U_A <= u_A | U_B = u_B) = dC(u_A,u_B)/du_B; signal when h1 < 0.05 (A cheap vs B) or h1 > 0.95.Fails when (read this): Copula family choice is a judgement; marginals must be fitted correctly first; both methods are studied on US data with 1990s-2000s costs — published returns decayed sharply after 2002 (Do & Faff 2010). - Index Inclusion / Rebalance Arbitrage: Passive funds must buy a stock on the effective date of index inclusion. Announcement-to-effective-date gap lets traders front-run that forced demand and unwind into the closing auction. The edge is the ratio of forced flow to the stock's normal liquidity.
Expected passive demand D = sum_funds (AUM_f * w_new). Flow pressure = D / ADV (days of volume). Expected impact ~ k * sqrt(D/ADV) * sigma (square-root law).Fails when (read this): Everyone knows the rules, so the anticipation trade is crowded and the 'pop' has shrunk over time; index committees can change methodology; the reversal can start before the effective date. - Cash-Futures Basis & Cost of Carry: A futures contract should cost the spot price plus the interest you save by not paying today, minus dividends you forgo. If the traded future is above that, buy spot / sell future and carry to expiry; if below, reverse. The basis (F-S) shrinks to zero on expiry.
F = S*(1 + r - q)^T (discrete) or F = S*e^{(r-q)T} (continuous). Basis = F - S. Implied carry r_imp = ln(F/S)/T. Arbitrage profit = F_mkt - F_fair - costs.Fails when (read this): Dividends cluster (ex-dates before expiry) so q is lumpy not smooth; your funding rate is not the repo rate; index arbitrage needs buying all 50 stocks in weight (tracking error, impact). - Calendar (Inter-Month) Spread in Futures: Two futures on the same index but different expiries should differ by one month of carry. When the far month trades richer than that, sell far / buy near; when cheaper, reverse. Same tick, same underlying — you are trading the funding rate the market implies between the two dates.
Fair spread = F_near * [(1+r-q)^(T2-T1) - 1]. Implied forward carry between expiries = ln(F_far/F_near)/(T2-T1).Fails when (read this): Far-month liquidity is thin so quoted spreads are wide; dividend timing differs between months; the spread can widen before it converges and margin is charged on both legs. - Roll (1984) Bid-Ask Spread Estimator: Trades bounce between bid and ask, so consecutive price changes are negatively autocorrelated. Roll showed that the size of that negative covariance reveals the effective spread even if you never see a quote.
Spread s = 2*sqrt( -Cov(dP_t, dP_{t-1}) ). Decomposition (Stoll 1989 / Glosten-Harris): s = order-processing + inventory + adverse-selection components; only the first is a pure 'bounce'.Fails when (read this): Positive serial covariance (trending ticks) makes the square root imaginary; assumes informationless order flow and constant spread; needs many trades. - Microprice (Imbalance-Weighted Mid): The midpoint ignores that a thin ask and a thick bid means the next move is probably up. Microprice tilts the mid toward the side with LESS size, giving a better guess of where the next trade prints.
Microprice = (P_ask * Q_bid + P_bid * Q_ask) / (Q_bid + Q_ask)Fails when (read this): Spoofed or iceberg size distorts Q; only level-1 depth is used; predictive power decays in milliseconds — a 200 ms retail feed sees it after the move. - Order Book Imbalance (OBI / OFI): Compares resting buy size with resting sell size near the touch. Cont et al showed short-horizon price changes are close to linear in the net order-flow imbalance, with slope inversely related to depth.
OBI = (Q_bid - Q_ask)/(Q_bid + Q_ask) in [-1, +1]. OFI_t = sum over events of (dBid volume) - (dAsk volume); dP ~ beta * OFI, beta ~ c / average depth.Fails when (read this): Hidden/iceberg orders, spoofing and layering (both illegal but present) invert the signal; effect horizon is far shorter than retail latency. - Avellaneda-Stoikov Market-Making Model: A market maker who is long wants to sell more than buy, so she shifts her 'personal mid' (reservation price) down in proportion to inventory, risk aversion, volatility and time left. Around that shifted mid she quotes a spread set by risk aversion and how fast orders arrive.
Reservation price r = s - q*gamma*sigma^2*(T-t). Optimal total spread delta = gamma*sigma^2*(T-t) + (2/gamma)*ln(1 + gamma/k). Quote bid = r - delta/2, ask = r + delta/2.Fails when (read this): Assumes Brownian mid with no drift (no adverse selection), exponential fill intensity, single asset, no latency; in reality informed flow hits your stale quotes — Guéant-Lehalle-Fernandez-Tapia extend it. - Glosten-Milgrom Adverse-Selection Spread: Even a zero-cost, risk-neutral dealer must quote a spread because some counterparties know more. The ask is the expected value of the asset GIVEN that someone chose to buy; the bid, given someone chose to sell. More informed traders = wider spread.
Two-state example: V = V_H or V_L with prob 1/2, fraction mu informed (trade with the truth), 1-mu noise (buy/sell 50-50). Ask = E[V | buy] = V_L + (V_H - V_L)*(1+mu)/2 ; Bid = V_L + (V_H - V_L)*(1-mu)/2 ; Spread = mu*(V_H - V_L).Fails when (read this): One-unit trades, no inventory cost, no order-processing cost; real spreads also carry Stoll's inventory and fixed-cost components. - Kyle's Lambda (Price Impact Coefficient): A market maker sets price as a linear function of net order flow: P = mu + lambda * (order flow). Lambda is the price you pay per unit of imbalance — the inverse of depth. Kyle solved it as an equilibrium between an insider, noise traders and the dealer.
P = mu + lambda * (x + u). Single-period Kyle: lambda = (1/2) * sigma_v / sigma_u, where sigma_v = sd of asset value, sigma_u = sd of noise order flow. Empirical: regress dP on signed volume, slope = lambda.Fails when (read this): Linear impact is an approximation — empirical impact is concave (square-root law); assumes normal distributions and a single insider. - Amihud Illiquidity Ratio: How much does price move per rupee traded? Average of |daily return| divided by daily rupee volume. Cheap to compute from daily data; a rough Kyle-lambda proxy.
ILLIQ = (1/D) * sum_d |r_d| / DVOL_d (often scaled x 10^6)Fails when (read this): Zero-volume days blow it up; conflates volatility with illiquidity; not comparable across very different price levels without scaling. - VPIN (Volume-Synchronised Probability of Informed Trading): Cuts the tape into equal-VOLUME buckets instead of time, classifies each bucket's volume as buy or sell (bulk classification), and averages |buy - sell| over the last n buckets. High VPIN = order flow is one-sided = informed/toxic = market makers about to widen or leave.
VPIN = sum_{i=1..n} |V_i^B - V_i^S| / (n * V), with V = bucket size, V_i^B = V * Phi((dP_i)/sigma_dP) (bulk volume classification).Fails when (read this): Andersen & Bondarenko (2014) showed VPIN is largely a volatility proxy and did not forecast the 2010 flash crash as claimed; bulk classification is crude; bucket size is arbitrary. - Almgren-Chriss Optimal Execution & Implementation Shortfall: Implementation shortfall (Perold) = the gap between a paper portfolio traded at the decision price and what you actually got. Almgren-Chriss chooses how fast to trade: fast = more impact, slow = more price risk; risk aversion lambda picks the point on that frontier, giving an exponential trading schedule.
IS = (Paper return) - (Actual return) = [delay cost + execution (impact) cost + opportunity cost + fees]. A-C: minimise E[cost] + lambda*Var[cost]; temporary impact h(v)=eta*v, permanent g(v)=gamma*v; optimal holdings x_j = X * sinh(kappa(T - t_j)) / sinh(kappa T), kappa^2 ~ lambda*sigma^2/eta. Expected cost of straight-line liquidation over T: (1/2)*gamma*X^2 + eta*X^2/T; variance = sigma^2 X^2 T/3.Fails when (read this): Assumes linear impact and arithmetic Brownian prices; ignores that others detect your schedule (predatory trading); parameters eta/gamma must be estimated from your own fills. - TWAP / VWAP / POV Execution Algorithms: TWAP: equal slices per time interval. VWAP: slices follow the historical intraday volume curve so you match the day's VWAP. POV: trade a fixed percentage of whatever volume prints, so you speed up when the market is busy. All three hide a big order inside normal flow.
TWAP slice_i = X/N. VWAP slice_i = X * (v_i / sum v) using historical intraday volume profile. POV slice = p * (market volume in interval), 0<p<1 (commonly 5-20%). Benchmarks: VWAP_day = sum(P_i*V_i)/sum(V_i).Fails when (read this): Historical volume curve fails on event days; POV chases volume spikes created by others; VWAP is a beatable benchmark that hides bad decisions inside an 'average'. - Maker-Taker Economics & Spread Capture: A scalper who posts limit orders 'makes' liquidity and earns the spread (and, on some exchanges, a rebate); one who hits the market 'takes' liquidity and pays the spread plus a fee. Expected P&L per round trip = spread captured - adverse selection - fees.
E[PnL_maker] = P(fill)*(s/2 - adverse_selection) + rebate - fee_maker ; E[PnL_taker] = edge - s/2 - fee_taker. Break-even edge for a taker = s/2 + all costs.Fails when (read this): Fill probability is not independent of direction — limit orders fill most when the price is moving against you (winner's curse); rebates do not exist on NSE/BSE. - Latency Arbitrage Mechanics: When a related price moves (say the future), resting quotes in the linked instrument (say the ETF or the option) are stale for a few microseconds. Whoever sees the move first and reaches the book first hits those quotes. Profit = stale-quote mispricing x probability of winning the race; the race is won by fibre, microwave and co-location, not by insight.
Expected sniping profit per event = (|dP_signal| - s/2) * P(win race) - fees ; P(win) ~ share of the fastest tier. Budish et al: the ES-SPY arbitrage duration fell from ~100 ms (2005) to ~10 ms (2011) but the count of opportunities did not fall - an arms race with no social gain.Fails when (read this): Speed bumps, frequent batch auctions, and randomised delays neutralise it; profits are winner-take-all so most entrants lose the fixed cost of the arms race. - Tick Size, Queue Position & Time Priority: NSE matches price-time priority: at a given price the earliest order fills first. If the tick is large relative to price, the spread is pinned at one tick and the queue is long, so being first in line is worth real money and a scalper's edge is the queue, not the price. If the tick is tiny, the queue is short but the spread is thin and quotes flicker.
Relative tick = tick / price. Queue value V(q) ~ (s/2 - AS) * P(fill | position q), decreasing in q. Expected wait ~ q / (arrival rate of marketable orders at that price - cancel rate ahead of you). Scalp gross = k ticks; needs k*tick > all costs per share.Fails when (read this): Queue estimates assume you see cancels ahead of you (needs TBT feed); priority resets when you modify quantity upward or price; hidden orders sit ahead invisibly. - India HFT Reality Check (2026): Not a formula — a fact sheet. True HFT in India runs from the exchange's co-location racks on the tick-by-tick (TBT) multicast feed with exchange-approved algos; it is a member-level, capital-intensive business. Retail API trading under the framework fully in force since 1 Apr 2026 is a separate, deliberately slower regime: a broker-mapped static IP, algo orders tagged with an exchange-issued ID, no direct exchange connectivity, and a threshold of 10 orders/second per client per exchange, above which the algo must be registered with the exchange through the broker.
Registration test: if orders/second > 10 (per exchange, per client) -> the algo must be registered with the exchange through the broker and gets a unique algo ID; at or below 10 no separate algo registration is needed, but static IP + exchange order tagging + broker routing + 2FA/OAuth + daily logout still apply. OTR charge: computed daily at member (broker) level on algo orders. The 2018 slab (NSE SURV/38122: <50 nil; 50-<250 Rs 0.02/order; 250-<500 Rs 0.10/order; >=500 Rs 0.10/order + 15-min order-entry restriction) was revised from 15 Jul 2020 (500-<1000 band Rs 0.15/order, higher bands added); use the exchange's current slab. Since 6 Apr 2026 (SEBI circular of 4 Feb 2026), equity-option algo orders within +/-40% of the premium LTP or +/-Rs 20 (whichever is higher) are exempt from the high-OTR penalty; other instruments keep the +/-0.75% LTP band, and designated market makers' market-making orders are left out of the OTR count.Fails when (read this): Rules are implemented through exchange and broker circulars that keep evolving (e.g. SEBI's 4 Feb 2026 OTR revision, which took effect on the exchanges from 6 Apr 2026); server-location and market-order restrictions are stated inconsistently across broker summaries — check the exchange's current circulars before relying on any figure here. - Black-Scholes-Merton Call & Put: Prices a European option assuming the underlying follows lognormal diffusion with constant volatility and a continuously-hedged riskless portfolio. Call = expected payoff under the risk-neutral measure, discounted. d1 carries the hedge ratio; d2 the risk-neutral probability of finishing in the money.
d1 = [ln(S/K) + (r - q + sigma^2/2)T] / (sigma*sqrt(T)); d2 = d1 - sigma*sqrt(T); C = S e^{-qT} N(d1) - K e^{-rT} N(d2); P = K e^{-rT} N(-d2) - S e^{-qT} N(-d1)Fails when (read this): Volatility is neither constant nor known; returns have fat tails and skew (hence the smile); American early exercise and discrete dividends are outside the base model; near expiry, gamma and pin risk make the smooth price meaningless. - Put-Call Parity: A call minus a put at the same strike and expiry equals a forward on the underlying struck at K. No model, no volatility — pure no-arbitrage. If it breaks, one of the four legs is mispriced or there is a cost (borrow, dividend, tax) you forgot.
C - P = S e^{-qT} - K e^{-rT} (equivalently C - P = (F - K) e^{-rT} with futures F). Synthetic long = C - P + K e^{-rT}. Implied rate r = -ln[(S - C + P)/K]/T.Fails when (read this): American options (early exercise) turn it into an inequality; discrete dividends before expiry; wide bid-ask on one leg makes the 'arbitrage' uncapturable. - Binomial Option Pricing (Cox-Ross-Rubinstein): Chops time into steps; each step the price goes up by u or down by d. A risk-neutral probability p makes the discounted expected price equal today's price. Work backwards from expiry payoffs to today. With many steps it converges to Black-Scholes; unlike BSM it handles early exercise naturally.
u = e^{sigma sqrt(dt)}, d = 1/u, p = (e^{r dt} - d)/(u - d). At each node V = e^{-r dt}[p V_up + (1-p) V_down] (American: max(that, intrinsic)).Fails when (read this): Constant volatility per step; oscillating convergence for barrier/discontinuous payoffs; slow for path-dependent options (use Monte Carlo). - The Greeks: Delta, Gamma, Theta, Vega, Rho: Delta: how much the option moves per point of spot (and the hedge ratio). Gamma: how fast delta changes — your 'curvature'. Theta: what you lose (or earn) per day just from time. Vega: sensitivity to a 1-point change in IV. Rho: sensitivity to rates — small for short-dated NIFTY options.
Delta_c = e^{-qT} N(d1), Delta_p = Delta_c - e^{-qT}. Gamma = e^{-qT} n(d1)/(S sigma sqrt T). Vega = S e^{-qT} n(d1) sqrt T (per 1.00 of sigma; /100 for per 1 vol point). Theta_c = -S e^{-qT} n(d1) sigma/(2 sqrt T) - r K e^{-rT} N(d2) + q S e^{-qT} N(d1) (per year; /365 per day). Rho_c = K T e^{-rT} N(d2) (/100 per 1%).Fails when (read this): Greeks are instantaneous partial derivatives — they change as soon as anything moves (that is what gamma, vanna, volga measure); theta is not earned linearly; on expiry day ATM gamma explodes and the linear picture is useless. - Implied Volatility via Newton-Raphson: There is no closed form for sigma given a price, so iterate: guess sigma, price the option, see how far you are, divide by vega (the slope) and correct. Converges in 3-5 steps for ATM options.
sigma_{n+1} = sigma_n - [BSM(sigma_n) - C_mkt] / Vega(sigma_n). Good start (Manaster-Koehler): sigma_0 = sqrt(2|ln(S/K) + rT| / T).Fails when (read this): Deep OTM/ITM options have tiny vega, so Newton overshoots or diverges (use bisection/Brent); prices below intrinsic or inside the bid-ask give no or many solutions; stale quotes give nonsense IV. - IV Rank vs IV Percentile: Both compare today's IV with the past year, but differently. IV Rank: where today sits between the 52-week low and high (a range position). IV Percentile: what fraction of past days had IV BELOW today's (a frequency). One outlier spike distorts Rank for a year; Percentile ignores it.
IVR = (IV_now - IV_52w_low) / (IV_52w_high - IV_52w_low) * 100. IVP = (# days in last 252 with IV < IV_now) / 252 * 100.Fails when (read this): Both are relative, not absolute: 'high' IV before a budget or election is high for a reason; a 1-year window can miss a regime change entirely. - Volatility Skew & Smile: BSM says one sigma for all strikes; the market disagrees. Index options show a skew: OTM puts carry higher IV than OTM calls because crashes are fast and everyone hedges the downside. Measure it as the IV gap between a 25-delta put and a 25-delta call (risk reversal) or the slope of IV vs strike.
Skew_25d = IV(25d put) - IV(25d call). Smile curvature (butterfly) = [IV(25d put) + IV(25d call)]/2 - IV(ATM). Slope proxy = [IV(K_low) - IV(K_high)] / [ln(K_high/K_low)].Fails when (read this): Skew from stale OTM quotes is noise; comparing skew across expiries needs delta-based not strike-based points; local-vol models fit today's smile but forecast its dynamics badly. - Expected Move (1-Sigma Range): The market's implied one-standard-deviation range to a date: spot x IV x sqrt(time). About 68% of outcomes should land inside it if IV is right. The ATM straddle price is a slightly smaller, cleaner version of the same thing (~0.8 x 1-sigma).
EM_1sd = S * IV * sqrt(T) (T in years; use days/365). Straddle approximation: EM ~ 0.8 * S * IV * sqrt(T) ~ ATM straddle price. Daily: S*IV/sqrt(252) or /sqrt(365).Fails when (read this): Assumes symmetric lognormal moves; ignores skew (downside range is wider in practice); IV changes over the period; gap events make sqrt(T) scaling wrong. - Max Pain, Put-Call Ratio & OI Reading Limits: Max pain: the strike at which the total intrinsic value of all open calls and puts is smallest — where option buyers as a group lose most. PCR: put OI (or volume) divided by call OI; high = many puts written or bought, read as either hedging or bullish contrarian. Both are popular and both are weaker than the folklore around them.
MaxPain = argmin_K { sum_i OI_call,i * max(K - K_i, 0) + sum_j OI_put,j * max(K_j - K, 0) }. PCR_OI = sum put OI / sum call OI; PCR_vol = put volume / call volume.Fails when (read this): OI counts contracts, not direction — a written put and a bought put look identical; max pain has weak, mostly-US evidence and NSE weekly expiries move the target daily; PCR extremes are regime-dependent and self-referential in crowded weekly options. - Dealer Gamma Exposure (GEX): Estimates how much dealers must buy or sell the underlying per 1% move to stay delta-neutral, given the open interest they are assumed to hold. Positive dealer gamma -> they sell rallies and buy dips (pinning, low realised vol). Negative -> they chase moves (whipsaws).
GEX = sum over strikes [ Gamma_call,i * OI_call,i * M - Gamma_put,i * OI_put,i * M ] * S^2 * 0.01 (convention: dealers long calls / short puts, i.e. customers buy puts and sell calls). Result in rupees of underlying to hedge per 1% move.Fails when (read this): The 'dealers short puts, long calls' assumption is a guess — OI has no side; retail-heavy markets break it; ignores futures/OTC hedges; single-stock GEX from thin chains is noise. - Delta Hedging & Gamma Scalping P&L: A delta-hedged long option earns 0.5 x Gamma x (move)^2 from re-hedging and pays theta each day. You profit when the underlying moves more than the break-even, i.e. when realised vol beats implied. The daily break-even move is sqrt(2 theta / gamma) — which is just S x IV x sqrt(dt).
Daily PnL ~ 0.5 * Gamma * (dS)^2 + Theta*dt (+ Vega*dIV). Break-even |dS| = sqrt(2*|Theta_daily| / Gamma). Over the life: PnL ~ 0.5 * sum Gamma_t S_t^2 (sigma_realised^2 - sigma_implied^2) dt.Fails when (read this): Discrete hedging adds noise and costs; theta is calendar-based but moves happen on trading days (weekend theta vs zero move); gaps make (dS)^2 huge and un-hedgeable — that is where short gamma dies. - Heston Stochastic Volatility Model (structure): Lets variance itself follow a mean-reverting random process, correlated with the spot. Negative correlation produces the index put skew; vol-of-vol produces the smile's curvature. Prices come from a semi-closed-form via characteristic functions, not a simple formula.
dS = mu S dt + sqrt(v) S dW1; dv = kappa (theta - v) dt + xi sqrt(v) dW2; corr(dW1, dW2) = rho. Feller condition 2 kappa theta > xi^2 keeps v > 0. Call price C = S P1 - K e^{-rT} P2 with P_j from Fourier inversion of the characteristic function.Fails when (read this): Cannot fit short-dated smiles well (needs jumps: Bates); calibration is unstable day to day; five parameters overfit a sparse NSE chain. - Realised vs Implied Volatility (Variance Risk Premium): Realised vol is what the underlying actually did (std dev of log returns, annualised). Implied vol is what options charged. Their gap, averaged over time, is the variance risk premium — the insurance margin option sellers collect and occasionally give back all at once.
RV = sqrt( (252/n) * sum_{i=1..n} (ln(P_i/P_{i-1}))^2 ) (zero-mean convention). VRP = IV^2 - RV^2 (variance units) or IV - RV (vol points). Parkinson RV = sqrt( (1/(4 n ln2)) * sum (ln(H/L))^2 * 252 ).Fails when (read this): Close-to-close RV misses intraday range (use Parkinson/Garman-Klass); 5-20 day windows are noisy; comparing 30-day IV with 5-day RV mixes horizons. - Payoff Algebra: Covered Call, Protective Put, Bull Call & Bear Put Spreads: Four building-block positions and their exact breakeven and limits. Premiums below are the theoretical BSM values from the 8-Jan-2025 NIFTY inputs so all four use one consistent chain.
Covered call (long S, short C_K): BE = S - C; max profit = (K - S) + C; max loss = S - C. Protective put (long S, long P_K): BE = S + P; max loss = S - K + P; unlimited upside. Bull call (long C_K1, short C_K2, K1<K2): debit D = C1 - C2; BE = K1 + D; max profit = (K2-K1) - D; max loss = D. Bear put (long P_K2, short P_K1, K2>K1): debit D = P2 - P1; BE = K2 - D; max profit = (K2-K1) - D; max loss = D.Fails when (read this): Payoff diagrams are at expiry only — before expiry, IV and time move the P&L; 'covered call' on an index needs futures or an ETF basket, not the index itself; early assignment does not apply to NSE European options but physical delivery of stock options does. - Payoff Algebra: Straddle, Strangle, Iron Condor, Butterfly: Straddle/strangle: long both sides, profit outside the breakevens. Iron condor: short strangle with bought wings — defined risk, profit inside. Butterfly: bet on a pin at the middle strike. Same chain as the BSM card so the numbers tie out.
Long straddle (C_K + P_K): cost = C+P; BE = K +/- (C+P). Long strangle (C_K2 + P_K1): cost = C+P; BE_low = K1 - cost, BE_high = K2 + cost. Iron condor (short P_K2, long P_K1, short C_K3, long C_K4): credit R; max profit R; max loss = width - R; BE = K2 - R and K3 + R. Long call butterfly (long C_K1, short 2 C_K2, long C_K3, equal widths w): debit D; max profit = w - D at K2; BE = K1 + D and K3 - D.Fails when (read this): Expiry-only pictures; pre-expiry the condor loses on a fast move even inside the wings (vega, gamma); butterflies need the pin to happen on the last day; straddles bleed theta while waiting for the move. - Payoff Algebra: Calendar Spread, Ratio Spread, Jade Lizard: Calendar: sell near-expiry, buy far-expiry at the same strike — profits from faster near-term theta and a rise in far IV; payoff at near expiry depends on the far option's remaining value, so there is no clean algebraic breakeven. Ratio spread (1x2): buy one, sell two further OTM — a credit or small debit with a profit peak at the short strike and open risk beyond. Jade lizard: short put + short call spread, credit greater than the call-spread width so there is no upside risk.
Calendar: debit D = C_far - C_near; value at near expiry = C_far(S_T, T_far - T_near, IV_far) - max(S_T - K,0) - D (numerical, no closed BE). Ratio 1x2 call (long C_K1, short 2 C_K2): net = 2C2 - C1 (credit if >0); max profit at K2 = (K2-K1) + net; upside BE = K2 + (K2-K1) + net; downside risk = -net if debit, none if credit. Jade lizard (short P_Kp, short C_Kc, long C_Kc+w): credit R = P + C_short - C_long; condition R > w -> no upside risk; downside BE = Kp - R; max profit R between Kp and Kc.Fails when (read this): Ratio spreads and jade lizards carry undefined risk on one side — a gap through the naked side is the whole story; calendars lose if far IV collapses or spot runs away from the strike; margin on naked legs is SPAN-based and rises with volatility exactly when the trade is losing. - India F&O Margins: SPAN + Exposure, ICMTM, Staggered Delivery Margins: Initial margin on NSE F&O = SPAN (portfolio worst-loss across 16 price/vol scenarios) + Exposure margin (a flat % of notional as extra buffer). ICMTM: intraday crystallised losses on futures are deducted from your available margin in real time, not at end of day. For stock derivatives, physical delivery margins phase in from four days before expiry on ITM options and futures.
IM = SPAN + Exposure. Exposure base rates (NSE Clearing): index futures 2% of notional, stock futures 3.5% (or higher per stock's volatility). ICMTM = sum of crystallised intraday losses on closed futures positions, netted against crystallised profits per client, adjusted against liquid assets real-time. Delivery margin on long ITM stock options / futures = k * (VaR + ELM + Adhoc) of the cash stock, k = 10% (E-4), 25% (E-3), 45% (E-2), 70% (E-1), 100% (E).Fails when (read this): SPAN parameters are re-set at least 5 times a day; margins jump on volatility spikes and SEBI/NSE add ad-hoc margins on event days — 'yesterday's margin' is not today's. - India 2026: Weekly Expiry Structure & STT on F&O: What actually expires when, and what the tax on each leg is, as of 2026. Weekly index options: one per exchange (SEBI Oct-2024 rule, effective 20 Nov 2024) — NIFTY on NSE, SENSEX on BSE. Expiry days: NSE Tuesday, BSE Thursday since 1 Sep 2025; NSE monthly contracts on the last Tuesday, BSE on the last Thursday. STT in this card's formula (classroom number, check your contract note): futures 0.05% of traded value (seller), option sale 0.15% of premium (seller), option exercise 0.15% of intrinsic value (buyer/exerciser).
STT_futures_sell = 0.0005 * price * qty. STT_option_sell = 0.0015 * premium * qty. STT_option_exercise = 0.0015 * intrinsic * qty (on ITM options settled at expiry). Rates: classroom number, check your contract note.Fails when (read this): Exchange expiry days have changed three times in two years (Thursday -> proposed Monday (deferred) -> Tuesday) and can change again; STT rates are set by the Finance Act each year — verify the current NSE circular and CBDT notification before teaching numbers. - Expectancy (per-trade edge): Tells you the average rupee (or R) you expect to make per trade after mixing win rate and average win/loss sizes. A positive number means the system has an edge on paper; a negative number means more trades only dig the hole deeper.
E = (W x AvgWin) - (L x AvgLoss), where W = win rate, L = 1 - WFails when (read this): Sample is small (under ~100 trades), costs are ignored, or the log mixes different setups and market regimes so the average hides a losing sub-system. - R-multiple: Expresses every trade's result as a multiple of the initial risk (1R = the rupees you would lose if the stop hit). A +2R trade made twice what it risked; a -1R trade lost exactly the planned amount; worse than -1R means the stop was not honoured.
R-multiple = (Exit - Entry) x Qty / Initial risk in Rs; Initial risk = |Entry - Stop| x QtyFails when (read this): Stops are moved after entry (R changes mid-trade), or when the initial risk is not written down before the trade so R is reconstructed after the fact. - Break-even win rate: Given a reward-to-risk ratio R, this is the minimum win rate at which expectancy is zero. Anything above it is profitable before costs.
W_be = 1 / (1 + R), where R = AvgWin / AvgLossFails when (read this): Costs are non-zero (the real threshold is higher), or when average win/loss are estimated from a handful of trades so R itself is unstable. - Risk of ruin: Probability that a fixed-bet strategy with a known edge loses its entire capital (or a defined ruin level) before growing. Even a positive-edge system can be ruined if capital is too few risk units.
For even-money payoff: RoR = ((1 - A) / (1 + A))^C, A = p - q (edge per bet), C = capital in units of bet sizeFails when (read this): Payoffs are not even (use Balsara's tables or simulation for uneven payoffs), trades are correlated, win rate is overestimated from a short backtest, or losses gap beyond the planned unit. - Kelly criterion: The fraction of capital that maximises the long-run growth rate of wealth when you know the odds and the win probability. Betting more than Kelly lowers growth and raises drawdown; betting double Kelly gives zero long-run growth.
f* = (b x p - q) / b = p - q/b, where b = net odds (win/loss ratio), p = win prob, q = 1-pFails when (read this): p and b are estimated with error (Kelly is extremely sensitive to overestimated edge), outcomes are not binary, bets overlap, or the trader cannot stomach the 50%+ drawdowns full Kelly implies. - Half / quarter Kelly: Bet a fixed fraction (1/2, 1/4) of the Kelly amount. Under the continuous approximation, growth rate at fraction k of Kelly is g(k) = (2k - k^2) x g_max: half Kelly keeps 75% of the maximum growth with roughly half the variance; quarter Kelly keeps about 44% with a quarter of the variance.
f = k x f*; growth ratio g(k)/g_max = 2k - k^2 (continuous-time approximation, Thorp 2006)Fails when (read this): The 2k - k^2 rule is a continuous, lognormal approximation; with fat-tailed trade outcomes the true growth curve is different and the optimum sits even lower. - Fixed-fractional position sizing: Risk a fixed percentage of current equity per trade; quantity falls out of the stop distance. Losses shrink the next bet automatically, which is what keeps a losing streak from becoming ruin.
Qty = (Equity x f) / |Entry - Stop|Fails when (read this): Stops are not honoured (gaps), multiple correlated positions each take f so portfolio risk is 5f, or the computed quantity rounds to zero on a small account. - Volatility-targeted sizing: Scale exposure so the position's expected volatility equals a target. High-vol instruments get less notional, low-vol get more, so every position contributes similar risk.
Weight = Target vol / Realised (or forecast) vol; Notional = Equity x WeightFails when (read this): Vol forecasts lag (realised vol spikes after the drop has happened), vol-of-vol is high, or de-leveraging by many vol-targeters at once amplifies the sell-off. - ATR-based stop: Put the stop a multiple of the Average True Range away from entry, so noisy instruments get wider stops and quiet ones tighter, and quantity follows from fixed-fractional sizing.
TR = max(H-L, |H-C_prev|, |L-C_prev|); ATR_n = Wilder-smoothed average of TR; Stop = Entry - k x ATR (long)Fails when (read this): ATR is computed on too few bars after a regime change, the instrument gaps (ATR does not protect against overnight gaps), or the multiplier is curve-fit to one backtest. - Value at Risk (historical, parametric, Monte Carlo): The loss that will not be exceeded with probability c over a horizon: 'we are 95% sure the day's loss is under Rs X'. Historical uses past returns directly; parametric assumes a normal distribution; Monte Carlo simulates from a chosen model.
Parametric: VaR_c = V x (z_c x sigma - mu). Historical: VaR_c = -(the (1-c) quantile of past P&L). Monte Carlo: same quantile from simulated P&L.Fails when (read this): Fat tails: real daily returns have far more 4-5 sigma days than a normal distribution allows, so parametric VaR understates tail loss. Historical VaR is blind to anything worse than its window. Neither says how bad the loss is once VaR is breached; that is CVaR's job. - CVaR / Expected Shortfall: The average loss on the days VaR is breached. Where VaR answers 'how bad is a bad day', CVaR answers 'how bad is the average of the worst days'. It is sub-additive (a coherent measure), which VaR is not.
CVaR_c = E[Loss | Loss >= VaR_c]. Normal case: ES_c = V x (sigma x phi(z_c)/(1-c) - mu)Fails when (read this): Estimated from a short history that contains no crash, or from a normal model when the book is short options (loss distribution is one-sided and unbounded). - Sharpe ratio: Excess return per unit of total volatility. The 1966 version divided by the sd of returns; the 1994 revision divides by the sd of the excess (differential) return itself, which is the form used today.
SR = (mean(R_p - R_f)) / sd(R_p - R_f); annualise by multiplying a periodic SR by sqrt(periods per year)Fails when (read this): Returns are skewed or fat-tailed (option selling shows high Sharpe until the tail hits), returns are autocorrelated (sqrt-time annualisation overstates), or the sample is short. Sharpe also punishes upside volatility equally with downside. - Sortino ratio: Like Sharpe, but the denominator only counts returns below a target (MAR). Upside swings are not treated as risk.
Sortino = (mean(R) - MAR) / DD, DD = sqrt( (1/n) x sum( min(0, R_i - MAR)^2 ) )Fails when (read this): Few sub-MAR observations make DD unstable; some vendors divide by the count of down periods only, which inflates the ratio; comparing Sortinos computed with different MARs. - Calmar ratio and MAR ratio: Annualised return divided by maximum drawdown. Calmar uses a trailing 36-month window; MAR ratio uses the full track record since inception. Both ask 'how much return per unit of worst pain'.
Calmar = CAGR (36 months) / |MaxDD (36 months)|; MAR = CAGR (inception) / |MaxDD (inception)|Fails when (read this): MaxDD is a single event, so the ratio jumps when the crash rolls out of the 36-month window; short tracks with no crash produce meaningless high values. - Omega ratio: Probability-weighted gains above a threshold divided by probability-weighted losses below it. Uses the whole distribution, so skew and kurtosis are inside the number rather than ignored.
Omega(L) = sum(max(R_i - L, 0)) / sum(max(L - R_i, 0)); Omega = 1 when mean return equals LFails when (read this): Sample lacks tail events; Omega at one threshold hides the shape at others (always plot Omega across L); highly sensitive to a single large loss. - Ulcer Index (and Ulcer Performance Index): Root-mean-square of percentage drawdowns from the running peak. Deep and long drawdowns both raise it, because every period spent below the high-water mark is squared and averaged. UPI = excess return / Ulcer Index.
UI = sqrt( (1/n) x sum( D_i^2 ) ), D_i = 100 x (V_i - Peak_i) / Peak_i (<= 0)Fails when (read this): Sampling frequency changes the number (daily vs monthly), so only compare like with like; it is backward-looking and says nothing about future drawdown. - Maximum drawdown, time to recovery and the recovery table: Max drawdown is the largest peak-to-trough fall in equity; time to recovery is how long the curve stays under the old high. The recovery table shows the gain needed to get back: it grows faster than the loss because you are compounding from a smaller base.
MDD = max over t of (Peak_t - V_t) / Peak_t; Gain needed to recover from loss L = L / (1 - L)Fails when (read this): Measured on month-end data only (intra-month troughs are missed), on a short history, or when leverage and margin calls force exit at the trough so the 'recovery' never happens for the leveraged account. - Beta, tracking error and information ratio: Beta measures how much the portfolio moves per unit move in the benchmark. Tracking error is the sd of active return (portfolio minus benchmark). Information ratio is mean active return per unit of tracking error: the Sharpe ratio of the bet against the index.
beta = Cov(R_p, R_b) / Var(R_b); TE = sd(R_p - R_b); IR = mean(R_p - R_b) / TE; annualise IR by sqrt(periods)Fails when (read this): Wrong benchmark (a mid-cap fund against NIFTY 50 shows fake alpha and huge TE), beta is unstable across regimes, or the sample is short. - Probabilistic Sharpe Ratio (PSR): Probability that the true Sharpe exceeds a benchmark SR*, given the observed Sharpe, sample length, skew and kurtosis. Negative skew and fat tails widen the Sharpe's standard error, so a 'high' Sharpe from a short, skewed track can have a low PSR.
PSR(SR*) = Phi( (SR_hat - SR*) x sqrt(T - 1) / sqrt(1 - g3 x SR_hat + ((g4 - 1)/4) x SR_hat^2) ), SR_hat in per-period units, g3 skew, g4 kurtosis (3 = normal)Fails when (read this): Returns are autocorrelated (the T-1 term assumes independence), the Sharpe was the best of many trials (then use DSR), or skew/kurtosis are themselves estimated from a tiny sample. - Deflated Sharpe Ratio (DSR): PSR with the benchmark raised to the Sharpe you would expect the best of N random trials to show. If you tried N parameter sets and picked the best, its Sharpe is inflated; DSR asks whether it beats that inflation.
SR0 = sqrt(V[SR_n]) x ( (1-gamma) x Z^-1(1 - 1/N) + gamma x Z^-1(1 - 1/(N e)) ), gamma = 0.5772 (Euler-Mascheroni); DSR = PSR(SR0)Fails when (read this): N is under-counted (every parameter tweak, every filter you tried, counts), trials are correlated (effective N is smaller but you must still estimate it), or V[SR] is guessed rather than measured across the trials. - Probability of Backtest Overfitting (PBO) via CSCV: Split the P&L matrix of all N trials into S blocks; for every combination of S/2 in-sample blocks, pick the best trial in-sample and record its rank out-of-sample. PBO is the fraction of combinations where the in-sample winner is below median out-of-sample. It measures how much your selection process itself overfits.
For each IS/OOS split c: w_c = rank_OOS(best_IS) / (N+1); logit_c = ln(w_c / (1 - w_c)); PBO = fraction of splits with logit_c < 0Fails when (read this): Trials are near-duplicates (effective N tiny), the P&L matrix is too short for S blocks, or blocks are so short that serial correlation leaks across them (pair with purging). - Stationarity and fractional differentiation: Prices are non-stationary (mean and variance drift), so most models need differencing; but full first-differencing (returns) throws away the price level's memory. Fractional differencing with 0 < d < 1 removes just enough trend to pass a stationarity test (e.g. ADF) while keeping most of the memory.
X~_t = sum_{k>=0} w_k X_{t-k}, with w_0 = 1 and w_k = -w_{k-1} x (d - k + 1) / k; pick the smallest d whose series passes ADF at 5%Fails when (read this): Series has structural breaks (ADF passes for the wrong reason), the weight window is truncated too early, or d is tuned on the test set. - Triple-barrier labelling: Label each event by which of three barriers the price path hits first: upper (profit-take, +1), lower (stop-loss, -1) or vertical (time-out, 0 or sign of return). Barriers scale with the recent volatility, so the label reflects how a real stop-managed trade would have ended.
Upper = P_0 x (1 + pt x sigma), Lower = P_0 x (1 - sl x sigma), Vertical at t_0 + h bars; label = first barrier touchedFails when (read this): Barriers are set in fixed points instead of vol units (labels change meaning across regimes), or overlapping events share the same price path (see sample uniqueness). - Meta-labelling: Keep your primary model (or discretionary rule) for the direction; train a second model whose label is simply 'did the primary signal make money (1) or not (0)'. The secondary model's probability then decides whether to take the trade and how big.
Primary: side_i in {+1,-1}. Secondary label: m_i = 1 if side_i x r_i > 0 else 0. Trade size = f(P(m_i = 1 | X_i))Fails when (read this): Secondary model sees features that were not available at signal time (look-ahead), or the primary model has no edge at all (meta-labelling filters noise into different noise). - Sample uniqueness and sample weights: When labels span overlapping windows, the samples are not independent: ten signals over the same week are closer to one observation than ten. Uniqueness measures how many labels share each bar; weights shrink over-counted samples so the model (and any bootstrap) is not fooled by copies.
c_t = number of labels whose window contains bar t; u_i = mean over the window of 1/c_t; weight w_i proportional to sum over window of |r_t| / c_tFails when (read this): Event windows are not tracked (label end times unknown), or weights are applied to the test set as well and then reported as 'accuracy'. - Purged and embargoed k-fold cross-validation: Plain k-fold shuffles or slices time so that a training label's window overlaps the test fold: the model 'sees' test outcomes through neighbouring labels. Purging drops training samples whose label window overlaps the test window; embargo additionally removes a buffer after the test fold so serially-correlated features cannot leak back.
Drop training sample i if [t_i0, t_i1] overlaps [test_start, test_end]; additionally drop i if t_i0 in (test_end, test_end + embargo]Fails when (read this): Feature engineering used future data upstream (purging cannot undo that), embargo is too short for slowly-decaying features (e.g. 200-day moving averages), or folds are shuffled randomly. - Walk-forward analysis: Optimise on an in-sample window, trade the next out-of-sample window with those parameters, roll both windows forward, repeat. Stitch the OOS segments into one equity curve. Walk-Forward Efficiency compares OOS to IS performance.
WFE = (annualised OOS return) / (annualised IS return); robust systems show WFE around 0.5-0.7 or betterFails when (read this): Window lengths are themselves optimised (another trial), the parameter surface is jagged so each re-optimisation jumps, or too few OOS segments exist to judge (fewer than ~8). - Feature importance: MDI and MDA: MDI (mean decrease impurity) sums how much each feature reduces impurity across tree splits: fast, in-sample, biased toward high-cardinality and substituted features. MDA (mean decrease accuracy) shuffles one feature at a time and measures the out-of-sample score drop: slower, honest, but shared information between correlated features is hidden.
MDI_j = mean over trees of sum over splits on j of impurity decrease (normalised to sum 1). MDA_j = (score_base - score_with_j_permuted) / score_base, computed on purged CV foldsFails when (read this): Features are strongly correlated (MDA under-reports each; cluster them first), MDI is read as if it were out-of-sample, or importance is computed on non-purged folds. - Bet sizing from predicted probability: Convert a classifier's probability into a size between 0 and 1: compute the z-statistic of p against 1/2 (or 1/K for K classes), pass it through the normal CDF, and map to [-1, 1] with the predicted side. Then average sizes of concurrent open bets and round to a step to avoid churn.
z = (p - 1/2) / sqrt(p(1-p)); m = side x (2 x Phi(z) - 1); then average over active bets and discretise to step sFails when (read this): Probabilities are not calibrated (a '65%' that wins 52% of the time), or the size is applied on top of leverage without a separate portfolio risk cap; bet sizing is not a substitute for fixed-fractional limits. - Why retail ML trading fails (checklist card): Five failure modes, each with the formula card that addresses it. (1) Non-stationarity: the data-generating process changes, so patterns learned in 2019 do not hold in 2022 -> fractional differentiation, walk-forward. (2) Low signal-to-noise: financial labels are mostly noise, so flexible models memorise it -> meta-labelling, sample weights, DSR. (3) Look-ahead bias: features or labels built with information not available at decision time -> purged/embargoed CV, triple-barrier with correct timestamps. (4) Survivorship bias: training on today's index constituents ignores the stocks that were removed -> point-in-time universes. (5) Transaction costs: a gross edge of a few basis points per trade disappears after brokerage, STT, slippage -> expectancy net of costs.
Net edge per trade = gross edge - (brokerage + STT + exchange fees + GST + stamp duty + slippage); if the model's gross edge is a few bps, net edge is negativeFails when (read this): This card 'fails' when it is used as a marketing checkbox: ticking 'we purged' without showing the code, the trial count N and the cost model is the same charlatanism Bailey et al. describe.
M5 · The Firms
Quant, algo and HFT firms: founded, founders, size, how they earn.
- Renaissance Technologies: Renaissance runs fully systematic, short-horizon models across equities, futures and currencies. Its flagship Medallion fund is closed to outsiders and owned by employees; it earns from thousands of small, statistically-edged trades with high turnover and leverage. Public funds RIEF and RIDA hold positions for longer and behave very differently from Medallion.
- D. E. Shaw & Co.: Started as a pure computational stat-arb shop above a communist bookstore; today a multi-strategy manager mixing systematic equity, macro, credit and private strategies. Revenue is management plus performance fees on outside capital, and India (Hyderabad) has been a major research and technology base for two decades.
- Two Sigma: Two Sigma treats investing as a data-science problem: large-scale machine learning over price, fundamental and alternative data, executed systematically. It earns fees on outside capital plus a market-making arm (Two Sigma Securities). Its founders' public dispute over governance (2023-24) led to new co-CEOs.
- AQR Capital Management: AQR is the academic-factor shop: value, momentum, carry, defensive and trend, run in liquid long-short and long-only portfolios. Fees are lower than hedge-fund norms because much of the product is 'alternative beta'. It publishes its research openly, and it went through a painful value drawdown in 2018-2020 before a strong rebound.
- Bridgewater Associates: Bridgewater codifies macro-economic cause-and-effect rules into systematic portfolios: Pure Alpha (active macro bets) and All Weather (risk-parity). Clients are mostly pensions and sovereign funds. Income is fees on very large institutional mandates; Dalio handed control to a new CEO and board in 2022-23.
- Man AHL (Man Group): AHL (Adam, Harding, Lueck) is the original systematic trend-follower, now the quant engine inside listed Man Group. It earns management and performance fees from institutional clients on diversified futures programmes, plus newer ML and multi-strategy books. Man Group's annual report is the public source for its assets.
- Winton Group: Harding left AHL and built Winton as a statistical-research house trading futures and equities. It earns fees on institutional capital. A cautionary tale: after publicly de-emphasising trend-following in 2018, it missed the 2020-22 trend rally and assets fell sharply from their mid-2010s peak before it re-embraced trend.
- PDT Partners: Process Driven Trading began in 1993 as a proprietary desk inside Morgan Stanley and became independent PDT Partners in 2012-13 after Volcker-era pressure on bank prop trading. It runs medium-frequency statistical equity strategies for outside investors and is famously small and secretive relative to its reputation.
- Millennium Management: Millennium is the archetypal multi-manager 'pod shop': hundreds of independent teams run tight risk limits under central risk control, and costs are passed through to investors. It profits from many uncorrelated small edges rather than one big call. Its pod model is what most new hedge funds now copy.
- Point72 Asset Management: Point72 is Steve Cohen's successor to SAC Capital: discretionary long-short equity pods plus the systematic arm Cubist and a macro book. It earns fees on outside money (re-opened to outsiders in 2018 after a family-office period). SAC's 2013 guilty plea is the regulatory backdrop every profile mentions.
- Cubist Systematic Strategies (Point72): Cubist is Point72's quant division, itself a multi-manager platform of independent systematic teams (equities, futures, options). Teams get data, execution and risk infrastructure and keep a share of their P&L. Its edge comes from platform scale rather than one master model.
- Citadel LLC: Citadel is a multi-strategy hedge fund (equities, fixed income/macro, commodities, credit, quant) with central risk management and pass-through expenses. It is distinct from Citadel Securities, the market maker, though both are Griffin's. LCH Investments ranked Citadel's 2022 result as the largest annual dollar gain by any hedge fund.
- Marshall Wace: Marshall Wace's signature is TOPS (Trade Optimised Portfolio System): it collects trade ideas from hundreds of sell-side analysts and brokers, scores their track records, and builds a systematic long-short equity book from the best of them. Alongside sit discretionary long-short funds. KKR took a stake in 2015.
- Squarepoint Capital: Squarepoint was spun out of Barclays' internal quant team when post-crisis rules pushed prop trading out of banks. It runs systematic strategies across equities, futures and credit for institutional clients, earning fees, and is known for hiring heavily from physics and CS PhDs.
- The Voleon Group: Voleon was built from day one on machine learning rather than economic hypotheses: models learn patterns from data with minimal human priors. It earns fees on institutional capital. Its founders are Berkeley statisticians, and the firm is often cited as the purest 'ML-first' hedge fund.
- WorldQuant: WorldQuant builds millions of small predictive signals ('alphas') and combines them into portfolios, primarily managing capital for Millennium. Its BRAIN platform crowdsources alphas from outside researchers worldwide — including a large Indian consultant base — paying for signals that survive testing. It is one of the biggest quant employers in India.
- Jane Street: Jane Street is a proprietary trading firm: no outside clients, all its own capital. It makes markets in ETFs, options, bonds and equities globally and profits from spreads and from keeping ETF prices in line with their baskets. It writes almost everything in OCaml. In India it became famous in 2025 when SEBI accused it of index manipulation on expiry days.
- Citadel Securities: Citadel Securities is the largest US retail equities wholesaler: brokers route customer orders to it, it fills them at or better than the public quote, and it keeps a slice of the spread (payment for order flow to brokers is the controversial part). It also makes markets in options, Treasuries and FX for institutions. Separate from Citadel the hedge fund.
- Jump Trading: Jump grew out of two CME floor traders and became a low-latency futures and equities firm, famous for owning microwave towers between Chicago and New Jersey to shave microseconds. Profit comes from being first to react across correlated markets and from market-making. Jump Crypto is its digital-asset arm, bruised by the 2022 Terra/Luna collapse.
- Hudson River Trading (HRT): HRT is an automated market-maker and quantitative trader across equities, futures, FX and crypto in dozens of countries, trading only its own capital. It earns spread and short-horizon alpha, and has pushed into longer-horizon 'mid-frequency' strategies. It opened a Mumbai office, making it one of the global HFTs with a local Indian presence.
- Virtu Financial (NASDAQ: VIRT): Virtu is a listed market maker, so its economics are public in its annual 10-K: 'adjusted net trading income' from making markets in 25,000+ instruments across 235+ venues, plus an execution-services business (from acquiring KCG in 2017 and ITG in 2019). It earns more when volatility and volumes rise. Read the 10-K's revenue and ADV tables for the actual numbers.
- XTX Markets: XTX is an algorithmic market maker in FX, equities, fixed income, commodities and crypto that explicitly does not compete on raw speed — it competes on prediction quality, using machine learning on huge datasets and a large GPU estate. It earns spread while holding inventory for seconds to hours. It has an office in Mumbai.
- DRW: DRW is a Chicago principal-trading firm founded by a CME options floor trader. It trades its own capital across futures, options, fixed income, energy and crypto (via Cumberland), earning from market-making and relative-value positions. It also invests in real estate and venture. It famously beat the CFTC in court over a 2011 interest-rate-swap-futures 'manipulation' case.
- IMC Trading: IMC began as two options floor traders on the Amsterdam exchange and is now a global electronic market maker, strongest in listed options and ETFs. It quotes continuously on exchanges and earns the spread while hedging risk. It is one of the Amsterdam trio (with Optiver and Flow Traders) and has a Mumbai office.
- Optiver: Optiver is a proprietary options and futures market maker: it posts two-sided quotes on thousands of listed derivatives and earns the spread, managing the resulting Greeks in real time. It trades only its own money and is known for its 80-question maths speed test in hiring and its Sydney and Chicago hubs.
- Flow Traders (Euronext: FLOW): Flow Traders is a listed ETF liquidity provider: it quotes ETFs on-exchange and off, hedges with the underlying basket, and earns 'net trading income' from the spread and creation/redemption arbitrage. Because it is public (Euronext, IPO 2015) its annual report shows how income spikes in volatile quarters and shrinks in calm ones.
- Susquehanna International Group (SIG): SIG is one of the largest options market makers in the US, trading its own capital. Its culture teaches decision-making under uncertainty through poker — new traders literally take poker classes. It also runs a venture arm (an early ByteDance/TikTok investor) and a sports-betting business. Profit is spread capture plus the edge from pricing options better than the crowd.
- Tower Research Capital: Tower is a multi-team HFT platform: independent trading groups share Tower's exchange connectivity, colocation and hardware, and keep a share of their P&L. It profits from ultra-short-horizon market-making and arbitrage in futures and equities worldwide. Its large Gurugram engineering centre made it one of the first global HFTs to build in India.
- Tradebot Systems: Tradebot is a small, private US equities HFT that trades only its own money, holding positions for seconds. Its founder Dave Cummings also founded the BATS exchange (2005) because he wanted a faster venue — BATS later merged into Cboe. Tradebot is the proof that HFT does not need Wall Street: it sits in Kansas City.
- Old Mission Capital: Old Mission is a Chicago proprietary market maker specialising in ETFs, options and fixed-income ETPs. It earns by quoting ETFs against their baskets and hedging instantly, and by being an authorised participant creating/redeeming ETF units. Smaller than Jane Street or Flow Traders but a real ETF-liquidity player.
- Akuna Capital: Akuna is a derivatives market maker founded by an ex-Optiver trader in Chicago. It quotes options across US and Asian venues and earns the spread while hedging Greeks, and it has expanded into crypto derivatives. It is a typical 'Amsterdam-school' options firm transplanted to Chicago.
- Wolverine Trading: Wolverine is a Chicago options market maker that grew from the CBOE floor into electronic quoting. Its affiliates also run an asset-management arm and an execution broker. Core profit is spread capture and volatility-surface pricing edge on listed options.
- GTS (Global Trading Systems): GTS is one of the largest Designated Market Makers on the NYSE floor (it bought Barclays' DMM unit in 2016), responsible for opening and closing auctions in hundreds of listed stocks. It also makes markets electronically in ETFs, Treasuries and FX. Profit comes from spread plus exchange incentives for DMM obligations.
- Radix Trading: Radix is a research-driven proprietary trading firm founded by ex-Citadel quants. It trades futures and equities using statistical models and describes its edge as 'science, not speed'. Small, private and quiet — included here as an example of the newer Chicago boutique quant-prop model.
- AlphaGrep Securities: AlphaGrep is one of India's largest home-grown quant proprietary firms: it trades its own capital across NSE/BSE cash, futures and options and in global markets from co-located servers. Income is market-making spread and short-horizon statistical arbitrage; it has expanded abroad and hires heavily from IITs.
- iRage Capital: iRage is one of the earliest Indian algorithmic prop firms, set up soon after NSE allowed DMA/co-location (2008-10). It trades its own capital in Indian equity derivatives using low-latency market-making and arbitrage. Its founders also created QuantInsti, the algo-trading education company, which is why the two are often confused.
- Graviton Research Capital: Graviton is a Gurugram proprietary HFT that trades Indian and global markets with its own capital, emphasising FPGA/low-latency engineering. Profit is spread and speed-based arbitrage. It is one of the most sought-after quant employers for Indian engineering graduates and is known for very high fresher pay.
- Quadeye Securities: Quadeye is a Gurugram quant prop firm trading its own capital in Indian and international exchanges with low-latency systems. It earns from market-making and short-horizon arbitrage; like Graviton it recruits mainly from IITs and is closed to outside investors.
- Estee Advisors: Estee is one of the few Indian quant firms that also manages outside money (PMS/AIF and an execution-algo business) rather than being pure prop. It earns fees on client capital plus proprietary quant trading in Indian equities and derivatives. Its founder ran a quant/analytics business in the US before returning.
- NK Securities Research: NK Securities is a Gurugram-based proprietary HFT and market-making firm active in Indian equity and derivatives markets from NSE co-location. It trades only its own capital. Small and private; it appears in campus-placement discussions more than in the press.
- Open Futures & Commodities: Open Futures is an Indian proprietary trading house active in equity derivatives, commodities and currency futures, one of the older domestic algo desks. It earns from market-making and arbitrage across NSE, BSE and MCX with its own capital.
- Dolat Capital / Dolat Algotech: Dolat is a Mumbai broking group whose listed arm (Dolat Algotech, formerly Dolat Investments) runs proprietary algorithmic trading in Indian equities and derivatives. Because it is listed on BSE/NSE, its annual report is one of the rare public windows into an Indian prop-algo P&L.
- Tower Research Capital India: The Indian arm of Tower Research Capital: engineering, research and trading from Gurugram, with trading servers in NSE co-location at BKC. It earns like the parent — spread and speed. It is the entity most often named when Indians discuss the NSE co-location controversy, although SEBI's principal 2019 co-location orders were directed at NSE and the broker OPG Securities.
- QuantInsti (EPAT): EDUCATION, not a fund. QuantInsti sells algo-trading courses (EPAT certificate, Quantra self-paced modules) and runs the Blueshift backtesting platform. It earns course fees, not trading profits. Founded by the iRage Capital founders, it is India's best-known quant-education brand; its certificates are useful but do not make anyone a licensed adviser or a guaranteed trader.
- DSP Quant Fund (DSP Mutual Fund): An example of a domestic AMC quant desk: DSP Quant Fund is an open-ended equity mutual fund whose stock selection follows a published rules-based model (quality, growth, value screens; exclusion of high-leverage and poor-governance names) rather than a fund manager's discretion. The AMC earns an expense ratio; retail investors can check its factsheet, AUM and expense ratio on the AMC and AMFI websites — a rare fully transparent 'quant' product in India.
M6 · The Map
Cities and countries that dominate algo trading, and why.
- Chicago (USA): Chicago is where futures were born (CBOT 1848, CME 1898) and where listed options were invented (CBOE 1973). Futures on the S&P 500, Treasuries, crude, corn and rates all price here, and CME Group's matching engine sits in Aurora, Illinois. Every firm that needs to be microseconds from the futures tape — Citadel Securities, DRW, Jump, Optiver, IMC, Akuna, Wolverine — keeps a Chicago base.
- New York / New Jersey (USA): Wall Street is the brand, but the US equity market physically lives in three New Jersey data centres: NYSE's matching engine in Mahwah, Nasdaq's in Carteret, and Equinix NY4/NY5 in Secaucus where Cboe/BATS, IEX and most dark pools sit. Being in all three, with the fastest links between them, is the whole game for US equities HFT.
- Chicago-New Jersey speed corridor (fibre to microwave) (USA): The most valuable 1,200 km in finance links CME's futures in Aurora to the equity engines in New Jersey, because S&P futures and SPY/ES arbitrage across it. In 2010 Spread Networks spent about $300m on a straight-line fibre route (roughly 13 ms round trip, cutting ~3 ms). Within two years microwave networks (McKay Brothers, Jump, others) beat it: light in glass travels at only about two-thirds of c, while radio through air moves at almost c, so a straight chain of towers is faster even though each hop adds a little delay. Later millimetre-wave and laser links squeezed the last microseconds; the practical limit is now ~8 ms round trip, close to the physics floor.
- London (LD4 Slough) (UK): London is the world's FX capital and Europe's derivatives hub: the London Stock Exchange, ICE Futures Europe (Brent, gilts) and the FX venues EBS/Refinitiv all match in or near Equinix LD4 in Slough, west of London, with LSE's engine in the same campus. Man AHL, Winton, Marshall Wace, XTX and Squarepoint were all built here.
- Amsterdam (Netherlands): Amsterdam has traded options since the world's first stock exchange (VOC shares, 1602) and Joseph de la Vega's 1688 'Confusion of Confusions' describes options and short-selling there. The modern reason is the 1978 European Options Exchange, whose floor bred Optiver (1986), IMC (1989) and later Flow Traders (2004). Post-Brexit, Cboe Europe and Turquoise moved EU share trading here, and Amsterdam overtook London in European equity turnover in 2021.
- Singapore (Singapore): Singapore is Asia's FX and commodities-derivatives centre (SGX iron ore, rubber, FX futures; a top-3 global FX centre by BIS surveys) and the regional HQ of choice for Western quant firms serving Asian hours — Jane Street, Citadel, Millennium, Squarepoint, HRT and Flow Traders all run large offices. For Indian readers its big historical role was SGX Nifty, the offshore Nifty futures contract that moved to GIFT City in July 2023.
- Hong Kong (China (SAR)): Hong Kong is the gateway to Chinese equities: HKEX lists most large mainland companies via H-shares and the Stock Connect pipes let global funds trade Shanghai/Shenzhen A-shares without an onshore licence. HKEX's Tseung Kwan O data centre hosts co-location. Every global market maker and multi-strategy fund keeps an Asia-equities desk here.
- Tokyo (Japan): Japan is the world's third-largest equity market and the Osaka Exchange's Nikkei 225 futures/options are among Asia's most liquid contracts. The Tokyo Stock Exchange's 'arrowhead' engine (2010) cut latency to milliseconds and opened the door to global HFT; Citadel Securities, Virtu, Optiver, Jane Street and Two Sigma all trade Japan from local entities.
- Shanghai (China): Mainland China is the largest retail-driven equity market on earth and its commodity futures exchanges (SHFE in Shanghai, DCE in Dalian, ZCE in Zhengzhou) are the world's most active. Domestic quant funds (Ubiquant, High-Flyer, Minghong, Lingjun) grew explosively after 2015; foreign firms (Two Sigma, Winton, Citadel Securities, Optiver) hold WFOE licences. High-Flyer's spin-off DeepSeek is the reason Chinese quant is now in every AI headline.
- Sydney (Australia): Sydney is the Asia-Pacific options hub because Optiver (1996), IMC and later Akuna and SIG built their APAC market-making out of the ASX options market — one of the world's largest single-stock options markets by contracts. Its time zone opens the global trading day, so APAC risk is run from here.
- Mumbai BKC (NSE co-location) (India): India's algorithmic trading physically happens in Exchange Plaza, Bandra-Kurla Complex — NSE's headquarters and co-location facility, where members rent rack space metres from the matching engine. NSE is the world's largest derivatives exchange by contracts traded, driven by weekly index options; that liquidity is why Jane Street, Citadel Securities, Optiver, IMC, HRT, XTX and Tower all built India desks alongside domestic firms like AlphaGrep, iRage and Dolat. BSE's engine sits a few kilometres away in Fort.
- GIFT City (IFSC, Gandhinagar) (India): GIFT Nifty on NSE IX: session 1 06:30-15:40 IST, session 2 16:35-02:45 IST next day, ~21 hours total, Mon-Fri, USD-settled
- Bengaluru / Hyderabad / Pune talent belt (India): Global quant firms have quietly built some of their largest engineering and research floors in south and west India: D. E. Shaw in Hyderabad since 1996, Goldman's and JPMorgan's quant-technology centres in Bengaluru, WorldQuant in Mumbai and Hyderabad, AQR, Millennium, Point72/Cubist and Citadel/Citadel Securities in Bengaluru or Gurugram, and captive tech for Morgan Stanley in Mumbai/Bengaluru. Pune hosts fintech and exchange-technology vendors. None of these are the firms' trading nerve centres; they are where models, data pipelines and risk systems get built.
M7 · Exams & Licences
What to actually study. India first, then global.
- NISM Series-I: Currency Derivatives Certification (NISM (SEBI's institute), Entry): Mandatory for approved users and sales staff of members in the currency derivatives segment; a clean primer on how FX futures/options are structured in India.
- NISM Series-V-A: Mutual Fund Distributors Certification (NISM, Entry): Required to get an AMFI Registration Number (ARN) and distribute mutual funds. The most-taken NISM exam.
- NISM Series-VIII: Equity Derivatives Certification (NISM, Entry): Mandatory for approved users and sales personnel of trading members in the equity derivatives segment. Covers F&O contract specs, margins, settlement, basic strategies and Greeks at a definitional level — the vocabulary every Indian F&O trader should have.
- NISM Series-X-A: Investment Adviser (Level 1) (NISM, Intermediate): First of the two exams required to register as a SEBI Investment Adviser (with Level 2). Covers personal finance, asset classes, risk profiling and the IA Regulations.
- NISM Series-X-B: Investment Adviser (Level 2) (NISM, Intermediate): Second mandatory exam for SEBI IA registration; case-study based on portfolio construction, insurance, retirement and estate planning.
- NISM Series-XIII: Common Derivatives Certification (NISM, Intermediate): One exam covering equity, currency and interest-rate derivatives — satisfies the requirement for all three segments at once for approved users.
- NISM Series-XV: Research Analyst Certification (NISM, Intermediate): Mandatory certification for SEBI Research Analyst registration — the licence 5 Circles itself holds (INH000020004). Covers fundamental and technical analysis basics, valuation, economics and the RA Regulations. From the 2024-25 amendments, black-box algo providers also need RA registration.
- NISM Series-XVI: Commodity Derivatives Certification (NISM, Entry): Required for approved users and sales staff in the commodity derivatives segment (MCX/NCDEX). Covers contract specs, delivery, warehousing, margins and the post-FMC-merger regulatory structure.
- NISM Series-XXI-A: Portfolio Management Services Distributors (NISM, Intermediate): Required for people distributing PMS products; covers the PMS Regulations, fee structures, performance reporting and investor suitability.
- SEBI Research Analyst (RA) registration (SEBI, Professional licence): The legal right to publish research and recommendations to clients or the public in India — including selling a black-box algo strategy under the 2026 retail-algo framework. Post the RA (Third Amendment) Regulations 2024 (effective 16 Dec 2024): a graduate or post-graduate degree in finance/commerce/economics/related fields (the earlier experience requirement was dropped), NISM Series-XV, and a client-count-based bank deposit (₹1 lakh to ₹10 lakh) in place of the old net-worth certificate.
- SEBI Investment Adviser (IA) registration (SEBI, Professional licence): The right to give personalised, fee-based investment advice in India. The IA (Second Amendment) Regulations 2024 parallel the RA overhaul: qualifying degree without the previous experience requirement, NISM Series-X-A and X-B, a client-count-based deposit replacing net worth, and a dual RA+IA registration path.
- CFA Charter (Chartered Financial Analyst) (CFA Institute, Advanced): The global benchmark for investment analysis and portfolio management: valuation, financial reporting, fixed income, derivatives, ethics. Recognised by Indian AMCs, PMS and research desks; SEBI accepts it as a qualification for RA/IA purposes.
- FRM (Financial Risk Manager) (GARP, Advanced): The standard credential for risk desks: VaR, credit and liquidity risk, derivatives valuation, Basel rules. Valued by Indian banks, NBFCs, exchanges and clearing corporations.
- CMT (Chartered Market Technician) (CMT Association, Advanced): The only widely recognised professional credential in technical analysis: trend, momentum, volatility, market breadth, behavioural finance, risk management and system testing. Has an active India chapter; useful for research roles that publish technical views.
- CQF (Certificate in Quantitative Finance) (Fitch Learning / CQF Institute, Advanced): A practitioner-taught programme in derivatives pricing, stochastic calculus, machine learning for finance, portfolio optimisation and Python. Good for a working trader who wants the maths behind the Greeks and the risk of ML overfitting.
- CAIA Charter (Chartered Alternative Investment Analyst) (CAIA Association, Advanced): Specialist credential for hedge funds, private equity, real assets, commodities and structured products. Relevant to Indian AIF (Category II/III) roles.
- PRM (Professional Risk Manager) (PRMIA, Advanced): An FRM alternative with strong coverage of mathematical foundations and case studies of risk failures (Barings, LTCM, etc.).
- CIPM (Certificate in Investment Performance Measurement) (CFA Institute, Intermediate): The specialist credential for measuring and reporting performance honestly: time-weighted vs money-weighted returns, attribution, GIPS standards. Directly useful for anyone who publishes a track record.
- FINRA Series 7 (General Securities Representative) (FINRA, Licensing): The core US licence to solicit and trade securities for clients at a broker-dealer. Relevant only if you work for a US-registered firm.
- FINRA Series 57 (Securities Trader) (FINRA, Licensing): The US licence for proprietary and agency equity/options traders at broker-dealers — market structure, order handling, Reg NMS, Reg SHO.
- Series 63 (Uniform Securities Agent State Law) (NASAA (administered by FINRA), Licensing): State-level registration to sell securities in the US; taken alongside Series 7.
- MIT OpenCourseWare 15.401 Finance Theory I (MIT, Foundation): Andrew Lo's Sloan course: present value, fixed income, equities, forwards/futures, options, portfolio theory, CAPM, market efficiency. Free lecture videos and notes — the same core as CFA Level I, taught by a researcher.
- MIT OpenCourseWare 18.S096 Topics in Mathematics with Applications in Finance (MIT, Advanced): Taught with Morgan Stanley practitioners: linear algebra, probability, regression, time series, volatility modelling, stochastic calculus, Black-Scholes derivation, portfolio theory, factor models. The closest free thing to a quant bootcamp.
- Aswath Damodaran's valuation classes and data pages (NYU Stern (Damodaran), Intermediate): Complete valuation and corporate-finance courses with slides, videos and spreadsheets, plus updated datasets (country risk premiums, industry betas, India ERP). The working reference for every DCF in this vault.
- QuantEcon lectures (Python and Julia) (QuantEcon (Sargent, Stachurski), Advanced): Rigorous, executable lectures on dynamic programming, Markov chains, linear state space models, filtering and asset pricing, with code. Built by a Nobel laureate (Sargent) and collaborators.
- NSE and BSE investor education (NSE Academy, BSE Institute, Investor Awareness) (NSE / BSE, Foundation): Exchange-run modules on market basics, F&O mechanics, margins, settlement and investor protection, in Indian context and often in Hindi. NSE Academy's NCFM modules are paid; the investor-awareness material is free.
- SEBI Investor website (investor.sebi.gov.in) and SCORES (SEBI, Foundation): SEBI's own investor education material, the registered-intermediary lookup (check any 'analyst' against the RA/IA register), the F&O risk disclosures, and SCORES for complaints. The first place to check before paying anyone for advice.
M8 · The Library
Books by category and level, one real takeaway each.
- Security Analysis, Benjamin Graham & David Dodd (1934): Before buying any stock, write its earning power (average of 5-7 years of profit) on paper and compare it with the market cap.
- The Intelligent Investor, Benjamin Graham (1949): Decide in writing whether you are a defensive or enterprising investor, and stick to that rulebook for one full year.
- Common Stocks and Uncommon Profits, Philip A. Fisher (1958): Pick one company you own and answer all 15 Fisher points in one page; where you cannot answer, you do not understand the business.
- One Up on Wall Street, Peter Lynch (with John Rothchild) (1989): Classify each stock you own as slow grower, stalwart, fast grower, cyclical, turnaround or asset play, and write a two-line 'story' for each.
- The Little Book That Beats the Market, Joel Greenblatt (2005): Build the two rankings on a NIFTY 500 sheet from screener.in data and see which 30 names come out on top; do not buy, just look.
- You Can Be a Stock Market Genius, Joel Greenblatt (1997): Track every NSE spin-off/demerger announced this year and note the parent and child price 90 days after listing.
- Margin of Safety, Seth A. Klarman (1991): For your next buy, write the liquidation value, the going-concern value and the price; only proceed if price is below the lower of the two.
- The Most Important Thing, Howard Marks (2011): Once a month, write one paragraph: where is the market in the cycle right now and what does that mean for your position size?
- Poor Charlie's Almanack, Charles T. Munger (ed. Peter Kaufman) (2005): Pick one of the 25 psychological tendencies each week and find one trade in your journal where it cost you money.
- Financial Shenanigans, Howard M. Schilit (1993): Run Schilit's receivables-vs-revenue growth check on your top holding for the last 5 years using the annual report.
- Quality of Earnings, Thornton L. O'glove (1987): Strip every 'exceptional item' out of your holding's last 3 years of profit and recompute the EPS.
- The Outsiders, William N. Thorndike Jr. (2012): For your holdings, list the last 5 years of capital allocation decisions (capex, buyback, dividend, acquisition) and grade each one.
- 100 Baggers, Christopher W. Mayer (2015): Compute how many years a stock needs at 20%, 25% and 30% CAGR to 100x, then ask if its market size allows that.
- Investment Valuation, Aswath Damodaran (1994): Build a two-stage DCF for one NIFTY 50 company in a spreadsheet and reconcile every input to the annual report.
- Narrative and Numbers, Aswath Damodaran (2017): Write a five-sentence story for a company you like, then check: does the story's total market size support the revenue your DCF assumes?
- The Dhandho Investor, Mohnish Pabrai (2007): For every trade or investment, write the downside in rupees first, then the upside; skip anything where downside is not small and known.
- Coffee Can Investing, Saurabh Mukherjea, Rakshit Ranjan & Pranab Uniyal (2018): Run the two-filter screen on screener.in yourself for the last 10 financial years and list every name that passes.
- The Unusual Billionaires, Saurabh Mukherjea (2016): Pick one company from the book and check on screener.in whether the ROCE and growth held up in the years after 2016.
- Diamonds in the Dust, Saurabh Mukherjea, Rakshit Ranjan & Salil Desai (2021): Apply the book's forensic checklist (cash conversion, contingent liabilities, auditor changes, related-party transactions) to one mid-cap you hold.
- Bulls, Bears and Other Beasts, Santosh Nair (2016): After each chapter, note the regulatory change that followed the episode (e.g., badla ban, rolling settlement, F&O launch) and the year.
- The Big Bull of Dalal Street, Neil Borate, Aprajita Sharma & Aditya Kondawar (2022): Write down the difference between RJ's 'trading book' and 'investment book' and decide which rules apply to each of your own positions.
- Value Investing and Behavioral Finance, Parag Parikh (2009): Look up the 1-year and 3-year post-listing returns of the last 10 large Indian IPOs and compare with the NIFTY over the same window.
- Stocks to Riches, Parag Parikh (2005): List the last three tips you acted on, who gave them, and what happened; then decide a rule for tips going forward.
- The Thoughtful Investor, Basant Maheshwari (2014): Write Maheshwari's promoter questions (skin in the game, dilution history, related-party deals) as a checklist and run one holding through it.
- The Scam: From Harshad Mehta to Ketan Parekh, Debashis Basu & Sucheta Dalal (1993): Draw the flow of money in a ready-forward deal on one page and mark exactly where the bank receipt fraud sat.
- Let's Talk Money, Monika Halan (2018): Set up the three-account money-box system this month and confirm 6 months of expenses sit in the emergency fund before any trading capital.
- Technical Analysis of the Financial Markets, John J. Murphy (1999): Mark trend, support and resistance by hand on 20 printed NIFTY daily charts before you use a single indicator.
- Japanese Candlestick Charting Techniques, Steve Nison (1991): Find 10 bullish engulfing candles on a stock's daily chart; note how many were at a prior support and what happened in 5 days for each group.
- Encyclopedia of Chart Patterns, Thomas N. Bulkowski (2000): Before trading any pattern, look up its failure rate in the book and write it in your trade plan next to your stop.
- How to Make Money in Stocks, William J. O'Neil (1988): Screen NSE for stocks with 25%+ quarterly EPS growth within 10% of 52-week high; keep a weekly list for 3 months without trading.
- Trade Like a Stock Market Wizard, Mark Minervini (2013): Compute your own expectancy from your last 30 trades (win rate × avg win − loss rate × avg loss) before reading the VCP chapter.
- Stan Weinstein's Secrets for Profiting in Bull and Bear Markets, Stan Weinstein (1988): Label every NIFTY 50 stock's current stage on a weekly chart with a 30-week MA; count how many are in Stage 2.
- Reminiscences of a Stock Operator, Edwin Lefèvre (1923): Copy out the ten rules Livermore breaks in the book and check which ones appear in your own journal.
- Trading in the Zone, Mark Douglas (2000): Do the '20-trade sample' exercise from the book: take 20 trades of one setup with identical size and no discretion, then judge only the sample.
- Market Wizards, Jack D. Schwager (1989): For each interviewee, write one line: what was their maximum risk per trade? Notice the range.
- The New Market Wizards, Jack D. Schwager (1992): Read the Druckenmiller interview and note his rule on when to bet big vs small; test it against your last 10 sized trades.
- Stock Market Wizards, Jack D. Schwager (2001): Compare the 2001 and 2003 editions' postscripts to see which methods survived the crash.
- Hedge Fund Market Wizards, Jack D. Schwager (2012): Read the Thorp interview and compute the Kelly fraction for one of your own setups from its win rate and payoff.
- Trading for a Living, Dr. Alexander Elder (1993): Adopt the 2% per-trade and 6% per-month loss limits for one quarter and log every time they stopped you.
- Diary of a Professional Commodity Trader, Peter L. Brandt (2011): Keep a Brandt-style public-format diary for 12 weeks: every trade, the chart, the reason, the result, no edits after the fact.
- Evidence-Based Technical Analysis, David Aronson (2006): Take one indicator rule you trust, backtest it on 10 years of NIFTY data, then apply White's Reality Check or a simple bootstrap to the result.
- New Concepts in Technical Trading Systems, J. Welles Wilder Jr. (1978): Compute 14-day RSI and ATR by hand for one stock on a sheet, then confirm against your charting platform.
- Bollinger on Bollinger Bands, John Bollinger (2001): Compute %b and BandWidth for 60 days of one stock and find the lowest-BandWidth day; note what happened after.
- Trading and Exchanges, Larry Harris (2003): Watch the NIFTY futures order book for 30 minutes and write down each order type you see used and by whom you think.
- Inside the Black Box, Rishi K. Narang (2009): Draw your own trading process as Narang's five boxes and write what sits in each; most retail systems have only the first box.
- Quantitative Trading, Ernest P. Chan (2008): Rebuild the book's mean-reversion example on two NSE stocks using survivorship-bias-free data (include delisted names).
- Algorithmic Trading: Winning Strategies and Their Rationale, Ernest P. Chan (2013): Run an ADF test on the spread of two NSE bank stocks over 2 years and compute the half-life of mean reversion.
- Algorithmic and High-Frequency Trading, Álvaro Cartea, Sebastian Jaimungal & José Penalva (2015): Implement the Almgren-Chriss schedule for selling 10,000 shares over a day and compare cost vs a naive VWAP split.
- Advances in Financial Machine Learning, Marcos López de Prado (2018): Apply purged k-fold CV to one strategy you have already backtested and compare the Sharpe with your original in-sample number.
- Machine Learning for Asset Managers, Marcos López de Prado (2020): Compute how many independent backtests it takes for the best one to show a Sharpe of 1.0 by luck alone, using the book's formula.
- Systematic Trading, Robert Carver (2015): Size one position by Carver's volatility-targeting rule (target annualised vol / instrument vol) instead of a fixed lot count.
- Active Portfolio Management, Richard C. Grinold & Ronald N. Kahn (1995): Estimate your own IC from 50 forecasts (rank correlation of predicted vs realised return) and compute the implied IR at your breadth.
- Expected Returns, Antti Ilmanen (2011): For each premium in the book, write one line on whether it has been shown to exist in Indian data, and find the paper if so.
- Efficiently Inefficient, Lasse Heje Pedersen (2015): Compute the annualised spread on one pending Indian merger (offer price vs current price, days to close) and compare it with the repo rate.
- What Works on Wall Street, James P. O'Shaughnessy (1996): Replicate one factor decile sort (e.g., P/S) on NIFTY 500 for 5 years using screener.in exports and compare with the book's US result.
- Quantitative Value, Wesley R. Gray & Tobias E. Carlisle (2012): Compute EBIT/EV for the 30 cheapest NIFTY 500 names, then remove any with a Beneish M-Score above −1.78.
- Deep Value, Tobias E. Carlisle (2014): Screen NSE for stocks trading below net current asset value; count them and note the sectors.
- Option Volatility & Pricing, Sheldon Natenberg (1988): For one NIFTY expiry, write the IV of ATM, 2 OTM and 2 ITM strikes and sketch the skew by hand; explain why it is not flat.
- Dynamic Hedging, Nassim Nicholas Taleb (1997): Delta-hedge a short straddle on paper for 10 days at end of day and record the P&L from gamma vs theta.
- Fooled by Randomness, Nassim Nicholas Taleb (2001): Simulate 1,000 coin-flip traders over 10 years in a spreadsheet and count how many show a 'genius' record by chance.
- The Black Swan, Nassim Nicholas Taleb (2007): Compute the NIFTY daily-return kurtosis over 20 years and count how many 4-sigma days actually occurred vs the normal prediction.
- The Volatility Surface, Jim Gatheral (2006): Fit an SVI curve to one NIFTY expiry's IVs and check that it does not allow butterfly arbitrage.
- Volatility Trading, Euan Sinclair (2008): Compute 20-day realised vol for NIFTY and compare with the ATM IV each day for a month; log the variance risk premium.
- A Man for All Markets, Edward O. Thorp (2017): Write down Thorp's three questions for any 'opportunity' (is there an edge, can it be sized, can the counterparty renege) and apply to your broker's latest product.
- Beat the Dealer, Edward O. Thorp (1962): Compute the Kelly bet for a 51/49 edge at even money and note how small it is.
- Beat the Market, Edward O. Thorp & Sheen T. Kassouf (1967): Price a warrant using the book's empirical curve and compare with a Black-Scholes value; note where they differ.
- The Man Who Solved the Market, Gregory Zuckerman (2019): List the three things Renaissance did that a retail trader cannot replicate, and the one thing (data hygiene) that they can.
- My Life as a Quant, Emanuel Derman (2004): Write Derman's 'Modelers' Hippocratic Oath' on the first page of your backtest notebook.
- When Genius Failed, Roger Lowenstein (2000): Compute what leverage would have taken your worst historical month to −100%, and write that number next to your margin.
- The Quants, Scott Patterson (2010): Check your strategies' correlation to each other during your three worst weeks; if it jumped, you have one strategy, not several.
- Dark Pools, Scott Patterson (2012): Read NSE's co-location and TBT feed pages and map each to the US equivalent in the book.
- Flash Boys, Michael Lewis (2014): Write one paragraph on whether the book's core complaint applies to NSE, given SEBI's co-location and TBT rules.
- More Money Than God, Sebastian Mallaby (2010): For each manager in the book, note the year their signature edge stopped working and why.
- The Big Short, Michael Lewis (2010): Estimate the carry cost Burry paid per year waiting to be right, as a percentage of his fund; note what 'early' cost.
- Liar's Poker, Michael Lewis (1989): List every product your broker has pitched you this year and the fee they earn on each.
- Against the Gods, Peter L. Bernstein (1996): Solve the St. Petersburg paradox on paper and write why expected value alone cannot size a bet.
- A Random Walk Down Wall Street, Burton G. Malkiel (1973): Compare your last 3 years of trading returns (after costs and taxes) with the NIFTY 50 index fund over the same period, honestly.
- The (Mis)Behavior of Markets, Benoît Mandelbrot & Richard L. Hudson (2004): Plot NIFTY daily returns as a histogram against a normal curve with the same mean and SD; count the tail excess.
- Thinking, Fast and Slow, Daniel Kahneman (2011): Find one anchoring error in your journal (holding because of your buy price) and write the rule that removes it.
- The Psychology of Money, Morgan Housel (2020): Write your own definition of 'enough' in rupees and what you would refuse to risk for more.
- The Essays of Warren Buffett, Warren Buffett (ed. Lawrence Cunningham) (1997): Compute 'owner earnings' (net income + D&A − maintenance capex) for one holding and compare with reported EPS.
- The Little Book of Common Sense Investing, John C. Bogle (2007): Compute the 20-year rupee difference between a 0.2% and a 2% expense ratio on Rs 10 lakh at 12% growth.
- The Warren Buffett Way, Robert G. Hagstrom (1994): Score one company on Hagstrom's 12 tenets, 0-2 each, and see where it fails.
- The Education of a Value Investor, Guy Spier (2014): Write your pre-investment checklist of 10 questions and use it on the next idea before looking at the price.
- The Little Book of Valuation, Aswath Damodaran (2011): Value one mature NIFTY company with a stable-growth DCF using only the annual report and a 10-year G-Sec yield.
- The Little Book of Behavioral Investing, James Montier (2010): Adopt the 'pre-mortem': before a trade, write why it failed as if it already had.
- The Alchemy of Finance, George Soros (1987): Find one Indian sector where rising stock prices improved the fundamentals (cheaper equity raising) and draw the loop.
- The Money Game, Adam Smith (George Goodman) (1968): Write one page on who you are as a trader: holding period, tolerance for drawdown, what you do when bored.
- How I Made $2,000,000 in the Stock Market, Nicolas Darvas (1960): Draw Darvas boxes on 5 NSE stocks at 52-week highs and note where the trailing stop would sit today.
- The Disciplined Trader, Mark Douglas (1990): For a week, log the emotion before each entry and exit in one word; count how many trades were 'to be right'.
- Pit Bull, Martin 'Buzzy' Schwartz (1998): Copy Schwartz's rule of taking a break after a big loss and define your own version in hours or days.
- Way of the Turtle, Curtis M. Faith (2007): Compute the Turtle 'unit' (1% of equity / ATR) for NIFTY futures at today's ATR and see how many lots it allows.
- Trend Following, Michael W. Covel (2004): Tabulate the drawdown and recovery time of one trend-following track record from the book against the S&P over 2008.
- The Complete TurtleTrader, Michael W. Covel (2007): Note which Turtles kept trading 20 years later and what they changed from the original rules.
- The Master Swing Trader, Alan S. Farley (2000): Pick one Farley setup and paper-trade it 20 times with a fixed stop before adding any other.
- Technical Analysis Explained, Martin J. Pring (1980): Chart NIFTY, 10-year G-Sec yield and crude on one page for 5 years and mark the turns that came in sequence.
- Intermarket Analysis, John J. Murphy (2004): Compute the 6-month rolling correlation of NIFTY with USD/INR and with Brent over the last 5 years.
- The Art and Science of Technical Analysis, Adam Grimes (2012): Test Grimes' 'pullback in an uptrend' on 100 NSE instances and record the excess return vs random entries.
- Studies in Tape Reading, Richard D. Wyckoff (as Rollo Tape) (1910): Label one NSE stock's last 12 months as accumulation, markup, distribution or markdown with volume evidence for each.
- Elliott Wave Principle, A. J. Frost & Robert R. Prechter Jr. (1978): Have two people independently count the waves on the same NIFTY chart and compare; note how often they agree.
- Options as a Strategic Investment, Lawrence G. McMillan (1980): Draw the payoff diagram of a NIFTY bull call spread and a bull put spread at the same strikes and compare max loss and margin.
- Options, Futures, and Other Derivatives, John C. Hull (1989): Price a NIFTY call with a 3-step binomial tree by hand and compare with Black-Scholes.
- Options Trading: The Hidden Reality, Charles M. Cottle (2006): Rewrite a NIFTY iron condor as its synthetic equivalents and confirm the Greeks match.
- Futures & Options Blueprint, ZebraLearn (ed.) (2022): Compute the total margin (SPAN + exposure) for one lot of a NIFTY short straddle using NSE's margin calculator and compare with the book's worked example.
- Absolute Power: Inside the NSE Co-location Scam, Sucheta Dalal & Debashis Basu (2023): Read SEBI's 2019 co-location order summary on sebi.gov.in and list the three findings in one line each.
- The Little Book of Market Wizards, Jack D. Schwager (2014): Pick the three lessons that most describe your own failures and write them on your screen frame.
- Unknown Market Wizards, Jack D. Schwager (2020): Note each trader's stated maximum drawdown and how they sized after it; compare with your own rule.
- Mastering the Market Cycle, Howard Marks (2018): Score the Indian market today on Marks' 'temperature' checklist (IPO volumes, credit spreads, retail participation) and keep the score monthly.
- Capital Returns, Edward Chancellor (ed.) (2015): For one Indian cyclical sector (cement, steel), plot industry capex vs ROCE over 10 years and mark the lag.
- Devil Take the Hindmost, Edward Chancellor (1999): Map the 1992 and 2000 Indian episodes onto Chancellor's bubble anatomy and list what matched.
- Manias, Panics, and Crashes, Charles P. Kindleberger (1978): Write the five Minsky stages with one Indian example for each from the last 30 years.
- Extraordinary Popular Delusions and the Madness of Crowds, Charles Mackay (1841): Read only the three financial chapters, then read a modern account of tulipmania and list what Mackay got wrong.
- Irrational Exuberance, Robert J. Shiller (2000): Compute a CAPE for the NIFTY 50 using 10 years of inflation-adjusted index earnings and compare with the current P/E.
- Misbehaving, Richard H. Thaler (2015): Measure your own disposition effect: average holding days of winners vs losers over the last 50 trades.
- The Hour Between Dog and Wolf, John Coates (2012): Add a 'streak counter' to your journal and a rule that halves size after 3 consecutive wins.
- The Daily Trading Coach, Brett N. Steenbarger (2009): Do lesson 1 today: write your best and worst trade of the week and the single difference in process.
- Building Winning Algorithmic Trading Systems, Kevin J. Davey (2014): Run a Monte Carlo reshuffle of your last 100 trade results and read off the 95th-percentile drawdown.
- Trading Systems and Methods, Perry J. Kaufman (1978): Implement Kaufman's Efficiency Ratio on NIFTY and compare the periods it flags as 'trending' with your eye.
- Python for Finance, Yves Hilpisch (2014): Reproduce the book's vectorised SMA-crossover backtest on NIFTY daily data with Indian costs (STT, brokerage) added.
- Pairs Trading: Quantitative Methods and Analysis, Ganapathy Vidyamurthy (2004): Test HDFC Bank vs ICICI Bank for cointegration over 3 years and estimate the spread's half-life.
- Finding Alphas, Igor Tulchinsky et al. (WorldQuant) (2015): Write one alpha as a single expression (e.g., rank of 5-day return reversal), neutralise by sector, and compute its IC on NIFTY 500.
- The Little Book of Stock Market Cycles, Jeffrey A. Hirsch (2012): Compute NIFTY's average monthly return by calendar month over 20 years and see which Almanac patterns survive.
- Common Sense on Mutual Funds, John C. Bogle (1999): Pull the SPIVA India report and note what share of large-cap active funds lagged the index over 10 years.
- The Zulu Principle, Jim Slater (1992): Screen NSE small caps for PEG below 1 with 3-year EPS growth above 20% and check how many pass a Piotroski filter.
- Accounting for Value, Stephen Penman (2011): Compute a two-year residual income value for one NIFTY company and compare with its market price and your DCF.
- The Intelligent Asset Allocator, William J. Bernstein (2000): Build a 3-asset (NIFTY, gilt fund, gold) portfolio on 15 years of data and compute the Sharpe of the mix vs each alone.
- The Simple Path to Wealth, J. L. Collins (2016): Write the 'F-you money' number for your household and how many months of trading losses it would survive.
M9 · Frauds & Scams
How each fraud worked, and which formula would have flagged it.
Every case carries a legal status tag. “Alleged” means an allegation, not a finding. Pending matters were re-verified on build day; the date is on the card.
- Enron (2001, USA): Debt and losses parked in off-balance-sheet partnerships (SPEs) run by its own CFO; mark-to-market accounting booked future profits today. CONVICTED (2026-10-31)
- WorldCom (2002, USA): Ordinary operating costs (line fees) were booked as capital assets, turning expenses into 'investments' and inflating profit. CONVICTED (2026-10-31)
- Bernie Madoff (2008, USA): Classic Ponzi: no trades were done; new investors' money paid old investors' withdrawals, statements were fabricated. CONVICTED (2026-10-31)
- Stanford International Bank (2009, USA / Antigua): Ponzi dressed as a bank: CD money went into illiquid personal ventures and new deposits paid old interest. CONVICTED (2026-10-31)
- Wirecard (2020, Germany): Profits attributed to 'third-party acquirers' in Asia and cash supposedly held in trustee accounts that did not exist. ALLEGED (2026-10-31)
- Luckin Coffee (2020, China / USA-listed): According to the SEC, the company fabricated more than $300 million of retail sales through related parties, inflated expenses to hide it, kept a fake operations database. SETTLED (no admission) (2026-10-31)
- Nikola / Trevor Milton (2020, USA): Founder's public claims about technology and orders that did not exist; a promo video showed a truck rolling downhill. CONVICTED — PARDONED (2026-10-31)
- LIBOR rigging (2012, UK / USA / global): Traders asked submitters to nudge daily LIBOR submissions up or down to suit derivative positions; during 2008 some banks low-balled to look healthier. CONVICTED (BANKS) (2026-10-31)
- FX 'Cartel' chat rooms (2015, USA / UK / global): Dealers at rival banks shared client orders in private chat rooms and traded ahead of the fix to move it in their favour. CONVICTED (BANKS) (2026-10-31)
- Volkswagen short squeeze (Porsche) (2008, Germany): When Porsche disclosed its stake in Oct 2008, the free float left was smaller than the outstanding short interest; shorts had to buy at any price and VW briefly became the world's most valuable company. NOT FRAUD — MECHANICS (2026-10-31)
- FTX / Alameda (2022, Bahamas / USA): Customer deposits were routed to the founder's trading firm through a code-level exemption; the hole was covered with FTX's own token as collateral. CONVICTED (2026-10-31)
- Bre-X Minerals (1997, Canada / Indonesia): Drill core samples were 'salted' with gold from elsewhere before assay, so every report showed richer and richer ore. ACQUITTED (2026-10-31)
- Theranos (2018, USA): Tests were secretly run on modified commercial analysers or gave unreliable results; investors and partners were shown false validation claims. CONVICTED (2026-10-31)
- Steinhoff (2017, South Africa / Germany): Fictitious and irregular transactions with related parties inflated profits and assets over years (per PwC-led investigation, 2019). CONVICTED (2026-10-31)
- Harshad Mehta securities scam (1992, India): Ready-forward deals between banks were routed through him; instead of real government securities, fake or unbacked Bank Receipts (BRs) from small banks moved bank money into the stock market. CONVICTED (2026-10-31)
- Ketan Parekh (K-10) and the 2025 front-running allegations (2001, India): 2001: circular trading funded by pay orders from co-operative banks (Madhavpura) and bank finance. 2025 (per SEBI interim order): a Singapore-based trader allegedly got advance word of a large US-based fund's upcoming trades and passed it to Parekh, whose associates in Kolkata traded ahead of them. INTERIM ORDER — FINAL ORDER PENDING (2026-10-31)
- Satyam Computer Services (2009, India): Years of inflated revenue and fictitious bank balances; fake invoices and forged bank confirmations, set out in the chairman's confession letter of 7 Jan 2009, which he later disowned in court; a special CBI court convicted him in 2015. CONVICTED (2026-10-31)
- NSEL payment default (2013, India): Buy a short-dated contract, sell a longer one; according to investigators, the spread was in effect a loan to borrowers whose warehouse stock largely did not exist. ALLEGED (2026-10-31)
- Saradha Group (2013, India (West Bengal)): Alleged Ponzi through a wide agent network in villages; investigators say new deposits paid old ones and money went into media and real estate. ALLEGED (2026-10-31)
- Rose Valley (2015, India (West Bengal / Odisha)): Collective investment dressed as hotel bookings; investigators allege a Ponzi, larger than Saradha by the ED's figures. ALLEGED (2026-10-31)
- PACL (Pearls Agrotech) (2014, India): An unregistered collective investment scheme (SEBI's finding, upheld by SAT); land mostly never allotted. ORDER UPHELD (2026-10-31)
- Sahara OFCDs (2012, India): A public issue disguised as private placement to avoid SEBI rules; investor records could not be verified. ORDER UPHELD (2026-10-31)
- PNB – Nirav Modi / Mehul Choksi LoU fraud (2018, India): According to PNB and the CBI, bank staff issued Letters of Undertaking over SWIFT without entering them in the core banking system, so overseas branches lent against guarantees PNB never recorded. ALLEGED (2026-10-31)
- IL&FS (2018, India): A maze of 300+ group entities hid leverage and evergreened loans; the SFIO alleged the old management and auditors concealed the group's true position. ALLEGED (2026-10-31)
- DHFL (2019, India): Per the CBI FIR, loans were sanctioned to shell entities and funds diverted; per the administrator's audit, 'Bandra book' fictitious retail loans. ALLEGED (2026-10-31)
- Yes Bank / Rana Kapoor (2020, India): Per CBI/ED: loans and investments to stressed groups (DHFL, ADA group) allegedly in exchange for kickbacks routed to family-controlled companies. ALLEGED (2026-10-31)
- PMC Bank – HDIL (2019, India): Management allegedly hid a huge single-group exposure (HDIL) behind thousands of dummy loan accounts so the regulator saw a diversified book. ALLEGED (2026-10-31)
- Karvy Stock Broking (2019, India): Client securities pledged using the power of attorney; borrowed money routed to group real-estate entities. PENALTY UPHELD (2026-10-31)
- NSE co-location and dark fibre (2015, India): Alleged preferential access: some brokers (notably OPG Securities) logged in first to tick-by-tick servers and secondary servers, getting data earlier; separately, a 'dark fibre' link gave a few members faster connectivity. SETTLED (no admission) (2026-10-31)
- Anugrah Stock & Broking (2020, India): Collected client funds under assured-return schemes and an unregistered PMS; client funds not segregated, securities pledged, accounts not settled. PENALTY IMPOSED — SEBI FINAL ORDER (2026-09-04)
- Sadhna Broadcast pump-and-dump (2023, India): Connected entities traded among themselves to lift a thin stock, then paid for misleading YouTube videos to pull in retail, and sold into that demand. FINAL ORDER — UNDER APPEAL (2026-10-31)
- Telegram/WhatsApp tip groups and SEBI impersonation (2025, India): Unsolicited group invites; fake trading apps and doctored testimonials; impersonation of real SEBI-registered intermediaries; pump-and-dump in small-caps coordinated through the groups; money collected to personal UPI accounts. ALLEGED (2026-10-31)
- Unregistered 'research' and advisory promising fixed returns (2025, India): Selling investment advice or research without SEBI IA/RA registration; charging fees against promised returns; SEBI passes orders directing refunds, market bans and penalties. SEBI FINAL ORDERS (2026-10-31)
- Jane Street – SEBI index manipulation case (2025, India / USA): SEBI's interim order (3 July 2025) alleges that between Jan 2023 and Mar 2025 the firm bought Bank Nifty constituents/futures in the morning to lift the index while holding much larger bearish options positions, then sold, profiting on the options ('intraday index manipulation' and 'marking the close'). Jane Street denies manipulation and says it was ordinary arbitrage/hedging. INTERIM ORDER — UNDER APPEAL (2026-10-31)
- Gensol Engineering / BluSmart (2025, India): SEBI alleges loans from IREDA/PFC meant for 6,400 EVs bought far fewer vehicles; funds moved through a dealer to promoter-linked entities and personal spending (including a luxury flat); lender 'conduct letters' given to rating agencies, saying the loans were being serviced, were allegedly forged. INTERIM ORDER — CONFIRMED, NOT FINAL (2026-10-31)
How to spot one today
- Anyone promising fixed, assured or 'guaranteed' returns from the market is breaking SEBI rules — walk away.
- 'Only today', 'last 5 seats', 'margin call, pay now' — urgency is the tool, not the opportunity.
- Screenshots and edited P&L are not a track record. Ask for a broker-verified statement over 12+ months, or assume none exists.
- Ask for the SEBI registration number (INA… for IA, INH… for RA) and check it yourself at the SEBI intermediary search: https://www.sebi.gov.in/sebiweb/other/OtherAction.do?doRecognised=yes — a certificate image is not proof.
- Never pay to a personal UPI/bank account or a 'trading app' link. Registered entities take fees in their registered name with an invoice.
- 'Proof' that lives only inside their group — screenshots, testimonials, celebrity photos — can be made in five minutes.
- You did not find them, they found you. Unsolicited WhatsApp/Telegram adds are the single most common entry point in SEBI's 2024-25 cautions.
- Your money should sit with a broker/AMC in your name and show in CDSL/NSDL statements — if someone 'manages' it in their account, it is not yours anymore.
- Complain at SEBI SCORES and, for money already lost, the cybercrime helpline 1930 — quickly; fund freezes work only in the first hours.
M10 · Reality Check
The honest, dated numbers on how retail actually performs.
- SEBI's F&O loss series, FY22 to FY26: 89% (FY22) → 93% (FY22-24) → 91% (FY25) → 87.7% (FY26) of individual traders in loss (2026-09-04). Four SEBI studies, one message: in every year measured, roughly nine in ten individual equity-derivatives traders finished the year with a net loss. Always attach the period to the number — '93%' describes FY22-24, not today. The FY26 improvement to 87.7% came with a 20% fall in active traders; the average net loss per trader went up, not down.
- Aggregate net loss by year: FY25 ₹1,05,603 cr (Jul 2025 study; restated ₹1,11,788 cr in Aug 2026) → FY26 ₹91,685 cr (2026-09-04). Individual traders' net losses in equity derivatives were over ₹1.8 lakh crore across FY22-24 (three years), ₹1,05,603 crore in FY25 alone (SEBI, July 2025), and ₹91,685 crore in FY26. SEBI's 2026 study restates FY25 as ₹1,11,788 crore when computing the roughly 18% decline; use that figure for the FY25-to-FY26 comparison and ₹1,05,603 crore when citing the 2025 study standalone. The cumulative FY22-26 figure of about ₹3.85 lakh crore is from the 5 Circles' internal research notes and press coverage, not yet checked against the SEBI PDF.
- FY26: fewer traders, bigger average loss: Active individual traders 98.1 lakh → 78.6 lakh (−20%); new entrants −40%; avg net loss per trader ≈ ₹1.17 lakh (2026-09-04). The headline fall in total losses is a participation story. SEBI reports active individual traders fell about 20% and new entrants about 40% in FY26, while the average net loss per trader rose to about ₹1.17 lakh from about ₹1.14 lakh. The people who stayed lost slightly more each. One outlet reported 87.5 lakh active traders; two others and the stated 20% decline support 78.6 lakh.
- Options are where the losses are: ≈92% of FY26 individual losses came from options; ≈88% of options traders lost vs ≈66% of futures traders (2026-09-04). SEBI's FY26 study attributes about 92% of individual net losses to options. The loss rate among options traders (about 88%) is far above futures traders (about 66%). Options are cheap to enter, so small accounts crowd in; time decay and transaction costs then do the rest. The futures-vs-options gap is the single most useful teaching fact in the series.
- Persistence is not learning: ≈90% of traders who lost two consecutive years and kept trading lost again the next year (2026-09-04). SEBI tracked the same individuals across years. Among those with losses in two consecutive years who continued trading, about 90% lost again in the following year. Experience alone did not change the outcome; the behaviour repeated. Higher turnover relative to capital was associated with higher loss rates, and younger, lower-income traders traded most intensely relative to their means.
- The house always gets paid: Individuals paid ≈₹1 lakh crore in transaction costs over FY22-26 (SEBI); ≈₹25,000 cr in FY26 alone (2026-09-04). Transaction costs — brokerage, exchange charges, STT, GST, stamp duty — are paid whether you win or lose. SEBI's FY26 study puts the five-year total for individuals near ₹1 lakh crore. In the FY22 study, transaction costs alone were a large share of loss-makers' net losses. A strategy with a small edge before costs is a losing strategy after them.
- The Indian fee chain, leg by leg: Brokerage + exchange txn charge + clearing + SEBI fee + STT + stamp duty + GST (2026-09-04). Every order pays up to seven parties. STT is statutory and changes with the Budget; the STT rates in the chart are a classroom number, check your contract note. SEBI charges a turnover fee, and GST applies on brokerage plus exchange/clearing charges. Exchange transaction charges and stamp duty vary by segment and change — read your broker's charges page and the contract note, and date-stamp any number you quote.
- Survivorship bias: the index you see is the survivors: Nifty 50 is rebalanced semi-annually; the current list is not the 1996 list (2026-09-04). Nifty 50 launched in 1996 (base 1,000 on 3 November 1995) and is reviewed every six months; losers are dropped and winners added. A backtest run on today's constituents over past years silently assumes you knew in 2005 which companies would survive to 2026. Use point-in-time constituent lists from NSE Indices' rebalance history; the exact count of original constituents still in the index is left unverified here.
- Why backtests lie: Overfitting, look-ahead, survivorship, ignored costs and slippage, multiple testing (2026-09-04). Five ways: (1) you tested 200 variants and kept the best — the Sharpe you see is the maximum of noise (Bailey & López de Prado's deflated Sharpe ratio corrects for this); (2) the data included information not available at the time; (3) delisted stocks are missing; (4) STT, slippage and impact were left out; (5) the sample is one regime. A backtest is a hypothesis, not evidence, until it survives out-of-sample and real costs.
- Barber & Odean (2000): trading is hazardous to your wealth: US households 1991-96: the most active 20% earned ≈6.5 percentage points a year less than the least active (2026-09-04). Using records of about 66,000 households at a US discount broker, Barber and Odean found that households that traded most earned an annual net return of about 11.4% against 18.5% for those that traded least, over 1991-96. Gross returns were similar; costs made the difference. Their conclusion — overconfidence drives excess trading, and trading costs eat returns — has held up across markets.
- Barber, Lee, Liu & Odean (2014): Taiwan day traders: Fewer than 1% of day traders earned predictably positive net returns (2026-09-04). Using the complete transaction record of the Taiwan Stock Exchange over 1992-2006, the authors found that the vast majority of day traders lose money after costs, and that less than 1% could predictably beat the market. A small skilled group exists and persists — which is exactly why everyone believes they are in it.
- Chague, De-Losso & Giovannetti (2020): Brazilian day traders: Of ≈1,600 who day-traded Bovespa futures for 300+ days, ≈97% lost money; ≈1.1% earned more than Brazil's minimum wage (2026-09-04). 'Day Trading for a Living?' followed everyone who began day-trading Brazilian index futures in 2013-2015. Among those who persisted for more than 300 sessions, about 97% lost money and only about 1.1% earned more than the minimum wage. Performance did not improve with experience. The mechanism is the same one SEBI found: persistence without a measurable edge.
- SEBI (Jul 2024): intraday cash traders, FY23: >70% of individual intraday cash traders lost; >500 trades/year group had an 80% loss rate (2026-09-04). Separate from the F&O series, SEBI's study of intraday cash-segment traders in FY23 found more than 70% lost money (average loss about ₹5,371), traders under 30 were 48% of the sample with a 76% loss rate, and very frequent traders lost most often. Loss-makers placed more trades on average than profit-makers.
- Expiry-day concentration in index options: Index-option turnover on expiry day: 70% (FY25) → 59% (FY26); within a week of expiry: 98% → 97% (2026-09-04). SEBI's data shows Indian index-option activity is overwhelmingly same-day and same-week. Concentration eased after the late-2024 measures (fewer weekly expiries, larger lots) but remains extreme; longer-dated contracts are marginal. Same-day options are the fastest feedback loop in finance, which is what makes them habit-forming.
M11 · Learning Paths
Beginner to quant: routes through everything above.
New Beginning (0-6 months)
Someone who has just opened a demat account or is about to.
- Get the household in order first
- Learn what a share and a market are
- Read a chart before trusting an indicator
- Risk before reward
- Paper-trade one setup, thirty times
- Know the mind you are trading with
- Decide: investor, trader, or both
Serious Learner (6-24 months)
Someone with a journal of 50+ real or paper trades and a written plan.
- Read the accounts like a forensic analyst
- Value a business two ways
- Combine fundamentals and charts into one system
- Understand options as volatility, not direction
- Test before you trust
- Learn from those who blew up
- Trade the plan, judge the sample
- Know the rules you trade under
Quant Track (24 months and beyond)
Someone comfortable with Python, basic statistics, and who has completed the Serious Learner backtesting rung.
- See the whole pipeline
- Mean reversion and cointegration
- Factors and where premia come from
- Stop fooling yourself
- Volatility as an asset
- Execution and microstructure
- Go live within the rules
M12 · Glossary
Every term on this page, in both languages.
- Alpha
- Return above what the market or a benchmark would have given for the same risk. Positive alpha is skill or luck; you need years of data to tell which.
- Beta
- How much a stock moves for a 1% move in the market. Beta 1.5 means roughly 1.5% for every 1% of NIFTY.
- Drawdown
- The fall from a peak to the next low in your capital, in percent. Maximum drawdown is the worst such fall.
- Sharpe ratio
- Excess return divided by volatility. Higher means more return per unit of risk taken, but it hides fat tails.
- Sortino ratio
- Like Sharpe, but only downside volatility counts as risk.
- Volatility
- How much prices swing, measured as the standard deviation of returns. Realised volatility is what happened; implied is what options expect.
- Implied volatility (IV)
- The volatility number that makes an option-pricing model match the market premium. It is the market's price of uncertainty.
- IV Rank vs IV Percentile
- IV Rank places today's IV between the 52-week low and high. IV Percentile is the share of days in the last year with lower IV. They are different numbers.
- Greeks
- Sensitivities of an option's price: delta (to spot), gamma (delta's change), theta (time), vega (volatility), rho (rates).
- Delta
- Change in option price for a one-point move in the underlying. Also read as a rough probability of expiring in the money.
- Gamma
- How fast delta changes as the underlying moves. Highest near the money and near expiry, which is why expiry days feel violent.
- Theta
- Daily loss in option value from the passing of time, all else equal. Sellers collect it, buyers pay it.
- Vega
- Change in option price for a one-point change in implied volatility.
- Open interest (OI)
- Number of outstanding contracts not yet closed. A rise means new positions; it does not say who is long or short.
- Put-call ratio (PCR)
- Put OI (or volume) divided by call OI. Used as a sentiment gauge; it has no fixed threshold and is easy to over-read.
- Max pain
- The strike at which the total value of open options would be lowest at expiry. A curiosity, not a magnet with proven pull.
- VWAP
- Volume-weighted average price for the session. Institutions benchmark execution against it; intraday traders use it as a fair-value line.
- Average True Range (ATR)
- Average size of a bar's true range over N periods, using Wilder's smoothing. A volatility unit for stops and position size.
- RSI
- Relative Strength Index: ratio of average gains to average losses over 14 periods, scaled 0-100, with Wilder's smoothing.
- MACD
- Difference between a 12- and 26-period EMA, with a 9-period signal line. A momentum and trend-change indicator.
- SMA / EMA
- Simple moving average weighs every day equally; exponential weighs recent days more, so it turns faster.
- Bollinger Bands
- A 20-period SMA with bands two standard deviations above and below. Band width tracks volatility, not direction.
- Support / resistance
- Price zones where buying or selling has repeatedly appeared. Zones, not lines; they break more often than tutorials admit.
- Trend
- A sequence of higher highs and higher lows (up) or lower highs and lower lows (down). Dow Theory's core object.
- Candlestick
- A bar showing open, high, low and close, with a body coloured by direction. Traced to Homma and the Dojima rice market.
- Breakout
- Price moving beyond a defined range or level, ideally on rising volume. Many fail, which is why stops exist.
- Mean reversion
- The tendency of a spread or price to return towards an average. Only tradeable when the average is stable and the half-life is short.
- Momentum
- Past winners keep winning for a while (typically 3-12 months). A documented factor with sharp crashes.
- Cointegration
- Two non-stationary series whose spread is stationary. The statistical basis of pairs trading; correlation alone is not enough.
- Stationarity
- A series whose mean and variance do not drift over time. Most prices are not stationary; returns and spreads may be.
- Half-life (mean reversion)
- Time for a spread to close half the gap to its mean, from an Ornstein-Uhlenbeck fit. Sets holding period and capital cost.
- Z-score
- Distance from the mean in standard deviations. Used for spread entries and for statistical screens.
- Hurst exponent
- Measures whether a series trends (H above 0.5), mean-reverts (below 0.5) or is a random walk (0.5).
- Kalman filter
- A recursive estimator that updates a hidden state (like a hedge ratio) as each new observation arrives.
- Basis / cost of carry
- Futures price minus spot. The no-arbitrage futures price is spot grown at the risk-free rate less dividend yield.
- Arbitrage
- Buying and selling the same value in two places to lock a riskless difference. Real arbitrage is tiny, fast and capital-heavy.
- Market maker
- A firm that continuously quotes bids and offers and earns the spread, managing inventory risk.
- Bid-ask spread
- Gap between the best buying and selling price. The market maker's income and the taker's first cost.
- Microprice
- A mid-price weighted by the sizes on the bid and ask. A better short-horizon fair value than the plain mid.
- Order book
- The live list of resting buy and sell orders by price. Its shape (imbalance, depth) is HFT's raw material.
- Slippage
- Difference between the price you expected and the price you got. Backtests that ignore it lie.
- Market impact
- How much your own order moves the price. Grows with order size relative to liquidity; Kyle's lambda measures it.
- Latency
- Delay between an event and your reaction to it. HFT competes in microseconds; retail in India operates in milliseconds to seconds.
- Co-location
- Renting rack space inside the exchange's datacentre so your server sits next to the matching engine. NSE offers it at BKC.
- Tick-by-tick (TBT) feed
- A data feed with every order-book change, not a snapshot. Available to co-located members; retail feeds are throttled snapshots.
- Order-to-trade ratio (OTR)
- Orders placed divided by trades executed. NSE charges penalties above set slabs to discourage quote stuffing.
- Static IP (SEBI retail algo framework)
- Under SEBI's retail algo rules operative from 1 Apr 2026, API orders must come from a whitelisted static IP registered with the broker.
- Algorithmic trading
- Orders generated and placed by a program under set rules. In India, exchange approval or a registered strategy ID applies above set order rates.
- High-frequency trading (HFT)
- Very fast, very short-horizon automated trading that earns tiny edges many times a day. Requires co-location and specialist infrastructure.
- Statistical arbitrage
- Systematic trading of many small mispricings between related securities, relying on averages rather than any one bet.
- Quant
- Someone who builds trading or investment decisions from data and models rather than judgement alone.
- Backtest
- Running a rule over past data to see what it would have done. Useful only with realistic costs and no peeking at the future.
- Overfitting
- Tuning a rule until it fits past noise. Looks brilliant in a backtest and fails live.
- Look-ahead bias
- Using information in a backtest that was not available at the time, such as a restated result or a close price before the close.
- Survivorship bias
- Testing only on stocks that still exist today, which quietly deletes every company that failed.
- Walk-forward analysis
- Fit on one window, test on the next, roll forward. Closer to how a strategy is actually run.
- Purged k-fold cross-validation
- Cross-validation that removes training samples overlapping the test window in time, plus an embargo, so labels do not leak.
- Deflated Sharpe Ratio
- A Sharpe ratio corrected for how many strategies were tried and for non-normal returns. Most backtested Sharpes shrink a lot.
- Probability of Backtest Overfitting
- The chance that the best in-sample strategy underperforms out of sample, estimated by combinatorial splits of the data.
- Kelly criterion
- The bet fraction that maximises long-run growth given edge and odds. Full Kelly is violent; practitioners use a fraction.
- Expectancy
- Average rupee result per trade: win rate times average win minus loss rate times average loss.
- R-multiple
- A trade's result expressed in units of the initial risk. A +2R trade made twice what it risked.
- Position sizing
- Deciding how much to buy so that a stop-loss hit costs a fixed fraction of capital. The one decision that keeps you alive.
- Risk of ruin
- Probability of losing enough capital that recovery is impossible, given win rate, payoff and bet size.
- Value at Risk (VaR)
- The loss you would not exceed on most days at a chosen confidence. Says nothing about the size of the rare day beyond it.
- Expected Shortfall (CVaR)
- Average loss on the days beyond the VaR threshold. Fixes VaR's blindness to tail size.
- Leverage
- Controlling more exposure than your capital, through margin or derivatives. Multiplies both gains and the speed of ruin.
- Margin call
- A demand to add funds when losses eat into required margin. Failure means the broker squares you off at the worst time.
- Short squeeze
- A rush by short sellers to buy back as price rises, pushing it further. Volkswagen 2008 and GameStop 2021 are the textbook cases.
- Hedge fund
- A private pooled fund that can go long and short and use leverage. Invented by Alfred Winslow Jones in 1949.
- Proprietary trading firm
- A firm that trades its own capital, not clients' money. Most HFT and market-making firms are prop firms.
- AUM
- Assets under management: money a fund manages for others. Not the same as a prop firm's trading capital or notional traded.
- Medallion (the point of it)
- Renaissance's internal fund, closed to outsiders since 2005. Its public funds behaved very differently, which is the lesson.
- DCF
- Discounted cash flow: value today equals future cash flows discounted at a required return. Every input is a guess; the discipline is in writing them down.
- Intrinsic value
- What a business is worth on its cash-generating ability, independent of the quoted price. An estimate, never a number you can prove.
- Margin of safety
- Buying well below your estimate of intrinsic value so that estimation errors do not ruin you. Graham's central idea.
- Free cash flow
- Cash from operations minus capital expenditure. The cash a business could hand to owners without shrinking.
- ROCE
- Return on capital employed: operating profit divided by equity plus debt. India's favourite quality ratio; Coffee Can uses 15%.
- ROE
- Net profit divided by shareholders' equity. DuPont splits it into margin, turnover and leverage so you can see why it is high.
- P/E ratio
- Price per share divided by earnings per share. How many years of current profit you pay for. Meaningless when earnings are near zero.
- PEG ratio
- P/E divided by expected earnings growth. Lynch's rough rule: near 1 is fair. Growth forecasts are the weak link.
- Book value
- Assets minus liabilities on the balance sheet. Reliable for banks and asset-heavy firms, misleading for software or brands.
- Accruals
- The gap between reported profit and cash flow. Persistently high accruals are the oldest earnings-quality warning.
- Piotroski F-Score
- Nine yes/no tests of profitability, leverage and efficiency. 8-9 is strong, 0-2 is weak.
- Altman Z-Score
- A weighted sum of five ratios that predicts financial distress. The emerging-market Z'' variant suits Indian companies best.
- Beneish M-Score
- Eight ratios that flag likely earnings manipulation. Satyam's reported numbers set it off before the confession.
- Promoter pledge
- Shares the promoter has pledged as loan collateral. High pledge plus a falling price can trigger forced selling. India-specific disclosure.
- Related-party transactions
- Business between a company and entities its promoters control. Large or unusual RPTs are a governance red flag.
- Factor
- A characteristic (value, size, momentum, quality, low beta) that explains return differences across many stocks over long periods.
- Efficient market hypothesis
- Prices reflect available information, so beating the market consistently is hard. Fama's three forms; still argued about.
- CAPM
- Expected return equals risk-free rate plus beta times the market premium. Elegant, widely used, empirically shaky.
- Efficient frontier
- The set of portfolios with the highest return for each level of risk. Markowitz, 1952.
- Risk parity
- Sizing assets so each contributes equal risk rather than equal capital. Bonds get leveraged; 2022 hurt it.
- Badla
- Old BSE system of carrying forward positions for a financing charge, a home-grown mix of margin and futures. Banned in 2001 after Ketan Parekh.
- Circuit limit / upper & lower circuit
- Daily price bands (2%, 5%, 10%, 20%) beyond which a stock cannot trade. At the band, one side of the book vanishes and you cannot exit.
- Market-wide circuit breaker
- Index falls of 10%, 15% and 20% halt the whole market for set periods. Triggered in March 2020.
- T+1 settlement
- Shares and money change hands one working day after the trade. India moved from T+2 in Jan 2023; a T+0 optional pilot followed.
- ASM / GSM lists
- Additional and Graded Surveillance Measures: exchange lists of stocks under watch, with extra margins or trade-for-trade settlement. Check before buying a runaway small-cap.
- Bulk deal vs block deal
- Bulk: a single client trades over 0.5% of a company's shares in a day, disclosed after. Block: a large negotiated trade in a special window at a price near market.
- Delivery percentage
- Share of a day's traded quantity actually delivered to demat rather than squared off intraday. High delivery suggests conviction, not just churn.
- F&O ban period
- When a stock's open interest crosses 95% of the market-wide position limit, new positions are barred until it falls below 80%.
- MWPL
- Market-wide position limit: the cap on total open interest in a stock's derivatives, set from its free float.
- SPAN margin
- The exchange's risk-based initial margin for F&O, computed from scenario losses across price and volatility moves.
- Exposure margin
- An extra margin over SPAN to cover mark-to-market risk beyond the scenarios.
- ICMTM
- Intraday crystallised mark-to-market: losses realised intraday are blocked from margin immediately rather than at day end.
- Physical settlement
- Stock F&O contracts held to expiry settle by delivering shares, not cash. Margins step up in the last four days for that reason.
- Weekly expiry
- Index options that expire every week. As of 2026 each exchange has one weekly index expiry (NSE Tuesday, BSE Thursday).
- 0DTE
- Options on their expiry day. Highest gamma, fastest theta, and the segment where most retail option losses concentrate.
- Freak trade
- A momentary print far from the prevailing price, usually from a market order hitting a thin book. Stop-loss orders get triggered by it.
- STT
- Securities Transaction Tax, charged on the value of trades. Rates on F&O rose from 1 Apr 2026; check the current schedule, not memory.
- Stamp duty
- A state levy on the buy side of trades, collected centrally since July 2020. Small per trade, large over a year of churn.
- Demat account
- Electronic holding of securities at NSDL or CDSL through a depository participant. Replaced paper certificates from 1996.
- Depository participant
- The broker or bank that operates your demat account on behalf of NSDL or CDSL.
- Clearing corporation
- The entity that guarantees settlement of every trade and holds the margins. NSE Clearing and ICCL in India.
- Research Analyst (SEBI RA)
- A SEBI-registered person or firm allowed to publish research and recommendations on securities, under the RA Regulations 2014 as amended.
- Investment Adviser (SEBI IA)
- A SEBI-registered adviser who gives personalised advice for a fee, under the IA Regulations 2013 as amended. Distinct from an RA.
- NISM
- National Institute of Securities Markets, SEBI's education arm. Its certification series are the entry exams for market roles in India.
- CMT
- Chartered Market Technician, the professional charter for technical analysis, run by the CMT Association.
- CFA
- Chartered Financial Analyst: a three-level curriculum in investment analysis. Builds analysts, not traders.
- Pump and dump
- Talking up a thinly traded stock through tips, channels or paid content, then selling into the buyers. Illegal; SEBI has ordered against it repeatedly.
- Front-running
- Trading ahead of a known large order or recommendation to profit from the move it will cause. A SEBI offence.
- Insider trading
- Trading on unpublished price-sensitive information. Regulated under SEBI's PIT Regulations 2015.
- Ponzi scheme
- Paying old investors with new investors' money while claiming a return. Collapses when inflows slow. Madoff, Saradha, PACL.
- Assured returns (red flag)
- Any promise of a fixed or guaranteed return from securities. No registered intermediary may make it. It is the first sign of a scam.
- Interim order (SEBI)
- A temporary SEBI direction issued before a full hearing, often impounding money or barring trading. It is not a finding of guilt.
- SAT
- Securities Appellate Tribunal, where SEBI orders are appealed. Its orders can go further to the Supreme Court.
- Repo 105
- An accounting move Lehman used to classify short-term borrowings as sales at quarter-end, shrinking reported leverage.
- Bank receipt (BR)
- A 1990s paper promise that securities would be delivered later. Forged or unbacked BRs were the engine of the 1992 scam.
- Ready-forward deal
- A sell-now, buy-back-later transaction in government securities, effectively a secured loan. Abused in the 1992 scam.
- Letter of undertaking (LoU)
- A bank's guarantee to another bank on a client's behalf, sent over SWIFT. Unauthorised LoUs were the PNB mechanism.
- Power of attorney (broker)
- A client's authorisation letting the broker move shares from demat. Karvy misused it to pledge client shares.
- Flash crash
- A very fast, very deep drop and rebound driven by automated selling into a vanishing order book. 6 May 2010 is the reference case.
- Portfolio insurance
- A 1980s strategy of selling futures as prices fall to replicate a put. Everyone selling together deepened Black Monday 1987.
- LTCM (as a lesson)
- Long-Term Capital Management: two Nobel laureates, 25x leverage, and a 1998 collapse when correlated spreads all widened together.
- Fat tails
- Extreme moves happen far more often than the normal curve predicts. Mandelbrot's point; every risk model that forgets it eventually blows up.
- Black swan
- A rare, high-impact event that looks obvious only afterwards. Taleb's term; the reason position size matters more than forecasts.
- Liquidity
- How much you can trade without moving the price. It disappears exactly when you need it most.
- Tick size
- The smallest allowed price increment. Sets the minimum spread and decides whether one-tick scalping can cover costs.
- Lot size
- The fixed number of units in one F&O contract, set by the exchange. It is why small accounts cannot size positions finely.
- GIFT City / IFSC
- Gujarat's international financial services centre with its own regulator (IFSCA), NSE IX and India INX, and near-22-hour derivatives trading.
- Scalping
- Very short trades aiming for a few ticks each, many times a day. Costs and spread decide viability more than skill does.
- Swing trading
- Holding for days to weeks to capture one move. Sits between intraday and positional.
- Paper trading
- Simulated trading with no money at risk. Teaches mechanics, not emotion; treat the results as an upper bound.
- Trade journal
- A written record of every trade with reason, size, result and R-multiple. The cheapest edge available to retail.