Skip to content
Menu
Finabase Learn

Articles · Great investors

Edward Thorp: from Beat the Dealer to quant hedge funds

How mathematician Edward Thorp beat blackjack, sized bets with the Kelly criterion, ran a convertible-hedge fund and spotted trouble at Madoff in 1991.

On this page
  1. The mathematician who beat blackjack
  2. Sizing the bet: Kelly
  3. From the casino to the market
  4. Princeton Newport Partners
  5. The Madoff warning
  6. What investors can take from Thorp
  7. How Finabase helps
  8. Frequently asked questions
  9. Summary
  10. Check your understanding

Edward O. Thorp, a mathematician born in 1932, did something casinos thought impossible and then did something Wall Street thought impossible. In 1962 his book Beat the Dealer showed that blackjack could be beaten by keeping track of the cards. A few years later he applied the same mix of probability, careful bet sizing and patience to the stock market, and from 1969 ran Princeton Newport Partners, one of the first hedge funds built on quantitative models.

Thorp matters to ordinary investors less for his returns than for three habits of mind he made famous: find a real, measurable edge before you bet; size every bet so that a run of bad luck cannot ruin you; and check claims that look too good to be true. This article covers his path from blackjack to arbitrage, the reported record of his fund, the hedge at the center of it, and his early doubts about Bernard Madoff.

The mathematician who beat blackjack

Thorp earned a doctorate in mathematics at UCLA and was teaching at MIT around 1960 when he worked out, on an IBM computer, how the player's advantage in blackjack changes as cards leave the deck. Most of the time the casino has a small edge. When the remaining cards are rich in tens and aces, the edge swings to the player. A player who tracks the cards and bets more in those moments can, over many hands, come out ahead.

He presented the result to the American Mathematical Society in January 1961 under the title "Fortune's Formula." At MIT he also worked with Claude Shannon, the founder of information theory, on a small wearable computer to predict roulette, which they tested in Las Vegas in 1961. Beat the Dealer followed in 1962 and became a bestseller. Casinos responded by adding decks, shuffling more often and barring counters, which Thorp took as confirmation that the edge was real.

Edward Thorp: from blackjack to hedge funds (dates from his memoir and publishers) 1961 Blackjack paper; roulette device with Shannon 1962 Beat the Dealer 1967 Beat the Market (with Kassouf) 1969 Princeton Newport Partners opens 1988 PNP winds down 1991 Flags Madoff for a client 2017 A Man for All Markets Sources: Thorp, Beat the Dealer (1962); Thorp and Kassouf, Beat the Market (1967); Thorp, A Man for All Markets (2017).
Key dates in Thorp's career, from his books and memoir.

Sizing the bet: Kelly

Knowing you have an edge is half the problem. The other half is how much to bet. Bet too little and the edge earns almost nothing. Bet too much and a normal losing streak wipes you out before the edge can pay. Thorp found the answer in a 1956 paper by John Kelly, a Bell Labs scientist and colleague of Shannon. Kelly's rule bets a fraction of capital that grows wealth fastest in the long run; for a simple even-money bet, the fraction equals the edge. The details and a worked example are in Kelly criterion: the fastest-growing bet size, and why to halve it.

Thorp used Kelly sizing at the blackjack table and later in his funds, but usually bet a fraction of the full Kelly amount. His reasoning, set out in his 1997 paper on the criterion, was that the inputs are never known exactly. Full Kelly is very volatile, and betting twice Kelly drives the expected growth rate to about zero. Underbetting costs a little growth; overbetting can cost everything. The same asymmetry is behind Position sizing: how many shares to buy for a set risk rules that limit the loss from any single trade.

From the casino to the market

In the mid-1960s, by then at the University of California, Irvine, Thorp turned to stock warrants. A warrant is a long-dated option issued by a company to buy its shares at a fixed price. Working with the economist Sheen Kassouf, he found that many warrants traded far above any reasonable value. Their 1967 book Beat the Market described the trade: sell the overpriced warrants short and buy the stock in a ratio that makes the position nearly indifferent to which way the stock moves. The profit comes as the warrant's price converges toward its value.

To decide which warrants were overpriced, Thorp needed a way to value them. He has written that by 1967 he was using a formula that is essentially the one Fischer Black and Myron Scholes published in 1973, though he never published his derivation, preferring to trade on it. Whatever the priority, the ideas behind How options are priced: the intuition behind the models were already his working tools.

Princeton Newport Partners

In November 1969 Thorp and Jay Regan launched Convertible Hedge Associates, soon renamed Princeton Newport Partners (PNP). Regan ran trading and operations from Princeton, New Jersey; Thorp ran research and models from Newport Beach, California. The fund's main business was convertible arbitrage: buying Convertible bonds: conversion price, premium and the bond floor, warrants and convertible preferred shares that its models said were cheap, and shorting the underlying stock to hedge. Later it added other relative-value trades, including early forms of statistical arbitrage.

The hedge in one picture

A convertible bond pays interest like a bond and can be swapped for a fixed number of shares. Its price therefore has a floor (its value as a bond) plus an option on the stock. If the convertible is cheap relative to that option, a fund can buy it and short enough stock to cancel the first-order effect of share price moves. That amount, the hedge ratio, is the option's delta (see Option delta: share equivalence, hedge ratio and its limits).

Illustration: a convertible bought and hedged with a short of 0.60 shares −$30 −$20 −$10 $0 +$10 +$20 +$30 $60 $70 $80 $90 $100 $110 $120 $130 $140 Convertible alone: gain or loss ($) Convertible + short stock hedge ($) +$3.81 if it falls to $70 +$2.86 if it rises to $130 Share price right after a sudden move (starts at $100)
Illustration with invented numbers: option value from Black-Scholes at 30% volatility, three years, no interest. Gains or losses right after a sudden move, before interest, borrowing costs, time decay and credit risk.

The picture is flattering on purpose: it shows only an instant move. In practice the option's time value decays each day if the stock goes nowhere, borrowing the shares costs money, and the bond can default. The fund's real profit came from buying convertibles below fair value and collecting interest, then managing those costs across hundreds of small positions, rebalancing the hedge as delta changed.

The reported record

Thorp reports in A Man for All Markets that from November 1969 to December 1988 PNP compounded at about 19.1% a year before fees and 15.1% after fees to its limited partners, against about 10.2% a year for the S&P 500 with dividends reinvested over the same span (the comparison as summarized by MoneyWeek). He also says the fund never had a losing year, including 1974, when US stocks fell about 26%. These are the manager's own figures. The partnership's audited statements are not public, so they cannot be checked independently, and a hedged fund's returns should be judged against its low volatility and its leverage, not only against a stock index.

How it ended

In December 1987 federal agents searched PNP's Princeton office, and in 1988 Regan and other Princeton employees were indicted under the racketeering (RICO) statute over trades alleged to have created false tax losses and broken securities rules. Thorp, in California, was not charged and has written that he knew nothing of the trades. Investors withdrew, and the partnership wound down at the end of 1988. In 1991 a federal appeals court vacated the tax-related convictions of several defendants, ruling that the jury should have been told about a good-faith defense, while upholding other counts (United States v. Regan). Thorp later ran a statistical arbitrage fund, Ridgeline Partners, in the 1990s until 2002.

The Madoff warning

In 1991 a consulting client asked Thorp to review its hedge fund investments. One manager stood out: steady monthly gains with almost no losses, from a strategy of buying stocks and hedging them with index options. By Thorp's account in his memoir, he obtained the client's trade confirmations and compared them with exchange records. For many of the options trades there was no matching trade on the exchange that day, and for others the reported size was larger than the total volume traded by everyone. He concluded that the trades had not happened, and the client pulled its money out. The manager was Bernard Madoff, whose fraud was exposed in December 2008.

Thorp did not go to regulators; the SEC was warned separately, and repeatedly, by others beginning years later. The episode is still a model of due diligence: rather than argue about whether the returns were plausible, he checked whether the trades existed.

What investors can take from Thorp

  • Measure the edge. Thorp did not trust intuition about blackjack or warrants; he computed the advantage. A trading idea without a measured edge is a guess (see Backtesting basics: biases, costs and an honest SPY test).
  • Bet in proportion. Kelly's rule and its fractional versions tie the size of a bet to the size and certainty of the edge. Most investors bet too much on their best idea.
  • Hedge what you do not want. His funds removed market direction so that their profit came only from the mispricing they had found.
  • Verify, do not admire. Smooth returns with no losing months are a reason to look harder.
  • Know when the edge is gone. He also argued that most investors, without a measured edge, do better in low-cost index funds, an echo of John Bogle and index investing: the cost matters hypothesis.

How Finabase helps

Finabase is not a hedge fund platform, but it has the pieces to practice Thorp's habits. A symbol's Options tab, for example SPY options, shows the chain with implied volatility, the input that decides whether an option looks cheap or dear. The Backtest page tests a rule on two years of history before any money is at risk, and paper trading lets you practice a position-sizing rule with play money and see how a streak of losses feels at different bet sizes.

Frequently asked questions

What did Edward Thorp discover?

He proved that card counting can give a blackjack player a mathematical edge, published it in Beat the Dealer in 1962, and later applied similar reasoning to warrants and convertible bonds, building one of the first quantitative hedge funds.

How did Edward Thorp use the Kelly criterion?

He sized bets as a fraction of capital in proportion to his edge, as Kelly's formula prescribes, but usually below the full Kelly amount because the edge is never known exactly and overbetting is much more damaging than underbetting.

How much did Princeton Newport Partners make?

By Thorp's own account, about 19.1% a year before fees and 15.1% after fees from November 1969 to December 1988, without a losing year. The figures come from the manager and have not been independently audited in public.

How did Thorp spot Madoff?

In 1991, reviewing a client's account, he compared Madoff's trade confirmations with exchange records and found that many reported options trades either did not exist or exceeded the volume traded that day. His client withdrew its money.

Summary

Edward Thorp showed that a careful mathematician could find small, real edges in games and markets and exploit them safely. He measured the edge, sized bets with a conservative version of the Kelly criterion, hedged away the risks he did not want, and checked suspicious claims against the facts. Princeton Newport Partners' reported record was strong, though it rests on his own figures, and the fund ended in a legal case that did not involve him. His most durable lessons are about process: an edge must be computed, bet in proportion and verified, not assumed.

Check your understanding

4 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. Why did card counting give Thorp an edge in blackjack?

    1. It let him predict the next card exactly
    2. The player’s advantage changes as cards leave the deck, so betting more when the remaining deck favors the player turns a small edge into a long-run profit
    3. Casinos paid card counters a bonus
    Show the answer

    B. The player’s advantage changes as cards leave the deck, so betting more when the remaining deck favors the player turns a small edge into a long-run profit Most hands still favor the house slightly. The profit comes from betting more in the minority of situations where the odds have swung to the player.

  2. Thorp usually bet less than the full Kelly amount. Why?

    1. Kelly maximizes losses
    2. The edge is never known exactly, and betting more than Kelly cuts long-run growth sharply while betting less costs only a little
    3. Casinos had table limits
    Show the answer

    B. The edge is never known exactly, and betting more than Kelly cuts long-run growth sharply while betting less costs only a little Overestimating the edge leads to overbetting, which can drive long-run growth toward zero or worse. A fraction of Kelly is a margin of safety.

  3. A fund buys a convertible bond and shorts 0.6 shares of the stock per bond. What is it mainly betting on?

    1. That the stock will rise
    2. That the convertible is cheap relative to its bond value plus conversion option, with the stock’s direction largely hedged
    3. That interest rates will fall to zero
    Show the answer

    B. That the convertible is cheap relative to its bond value plus conversion option, with the stock’s direction largely hedged The short stock cancels most of the effect of small price moves. The profit comes from the mispricing, interest and large moves either way, minus borrowing costs and time decay.

  4. How did Thorp test whether Madoff’s returns were real in 1991?

    1. He compared the returns with the S&P 500
    2. He checked the reported options trades against exchange records and found many that did not exist or exceeded total volume
    3. He asked Madoff for references
    Show the answer

    B. He checked the reported options trades against exchange records and found many that did not exist or exceeded total volume Rather than debate whether the returns were plausible, he verified whether the trades had happened at all.

Sources and further reading

  1. Thorp — Fortune’s formula: a winning strategy for blackjack, Notices of the American Mathematical Society, 1960 (abstract; talk given January 1961)
  2. Thorp — Beat the Dealer, Blaisdell, 1962; revised edition, Random House, 1966
  3. Thorp and Kassouf — Beat the Market: A Scientific Stock Market System, Random House, 1967
  4. Kelly — A new interpretation of information rate, Bell System Technical Journal 35(4), 1956
  5. Thorp — The Kelly criterion in blackjack, sports betting and the stock market, 1997 (in Handbook of Asset and Liability Management, 2006)
  6. Thorp — A Man for All Markets: From Las Vegas to Wall Street, Random House, 2017
  7. MoneyWeek — Edward Thorp: the world’s greatest investors (Princeton Newport record)
  8. Poundstone — Fortune’s Formula, Hill and Wang, 2005
  9. United States v. Regan, 937 F.2d 823 (2d Cir. 1991)