Ed Thorp and Claude Shannon: beating markets through unusual skill

Executive overview

Most academics insisted markets were perfectly efficient and no one could beat them. Shannon and Thorp proved them wrong — independently, using rigorous math — first in casinos, then on Wall Street.

Shannon invented information theory and later outperformed 1,025 of 1,026 mutual funds using a concentrated buy-and-hold approach. Thorp invented card counting, built a wearable roulette computer, and ran a hedge fund returning 15.1% annually over 19 years vs. the S&P's 8%.

Both relied on the Kelly Criterion: bet a fixed fraction of your bankroll proportional to your edge. Never risk ruin. The contrast with LTCM — which ignored Kelly and imploded — illustrates why this principle matters.

Only compete where you have a genuine edge; never accept any risk of losing everything.

Claude Shannon: from information theory to investing

• Invented information theory in the late 1940s — the mathematical foundation of computers, the internet, and all digital media
• Described as doing what almost no one had done since the Renaissance: single-handedly founding an important new science
• Worked alone, slept when he felt like it, and spent hours thinking without external input — a habit he shared with Thomas Watson of IBM
• Applied insights from wartime cryptography (Project X with Alan Turing) directly to information theory; the two lines of inquiry were nearly inseparable
• Organized the first major academic AI conference in 1956, decades before the term entered mainstream use
• Began an intensive stock market study in the late 1950s, filling three shelves with ~100 books on investing

How Shannon invested

• Rejected technical analysis outright — price charts are "a very noisy reproduction of the important data"
• Focused on earnings growth plotted on logarithmic paper, drawing trend lines into the future
• Visited technology startups, met management, and became an active Teledyne board member
• Held 81% of his portfolio in a single stock (Teledyne); his three largest positions made up 98% of the portfolio
• Outperformed all but one of 1,026 mutual funds tracked by Barron's in 1986
• Achieved a 28% annual return from the late 1950s through 1986 — slightly above Buffett's 27% over a comparable period
• Operated with his wife and a second-hand Apple II computer, not a large team

Ed Thorp: from physics to blackjack to Wall Street

• Showed exceptional quantitative instincts from childhood — could total grocery bills faster than an adding machine
• Obsessed with maximising skill per hour: measured casino winnings on a per-hour basis and applied the same lens to investing
• Recognised that beating roulette was possible whether wheels were perfect (physics predicts the ball) or flawed (biased numbers)
• Taught himself Fortran to run blackjack simulations on MIT's mainframe IBM 704
• Submitted his blackjack paper through Claude Shannon — the only National Academy member at MIT who was a mathematician
• Collaborated with Shannon to build a wearable roulette computer (cigarette-pack size, 12 transistors, toe switches, earpiece); wiring failures ended the Las Vegas test
• Wrote Beat the Dealer, which sold over a million copies and established card counting

Applying casino math to the stock market

• Stopped casino play after casinos drugged his drink; redirected the same edge-seeking framework to equities
• Lost money speculating on silver early on; concluded: "you are unlikely to get an edge out of what you see in the news"
• Developed delta-hedging of warrants at UC Irvine, turning $40,000 into $100,000 by 1967
• Founded Princeton-Newport Partners with a bi-coastal remote structure — Thorp handled the math in California, Reagan handled investor relations on the East Coast
• Survived Black Monday (October 1987) essentially flat for the month while most funds fell 20%+; returned 34% for the year
• Princeton-Newport's 19-year record: a dollar invested grew to ~$15; 15.1% compound annual return vs. 8% for the S&P

The Kelly Criterion

• Core principle: bet a fixed fraction of your bankroll proportional to your edge; as the bankroll shrinks, bets shrink — you can never be wiped out
• Even unlikely events must eventually occur; anyone who accepts a small risk of ruin will eventually be ruined
• Buffett, Munger, Shannon, and Thorp all used or understood Kelly; Paul Samuelson publicly dismissed it while privately buying Berkshire Hathaway stock
• Thorp's rule: if you don't have an edge, don't bet and don't compete

LTCM: the cost of ignoring Kelly

• Long-Term Capital Management was founded by Merriweather and staffed with Nobel laureates (Merton, Scholes) who believed in efficient markets but exploited tiny temporary mispricings with extreme leverage
• Initial leverage: $29 borrowed for every $1 of investor equity; ballooned to ~60x as losses mounted
• When things went wrong, they increased bets — the opposite of the Kelly system
• Required a US Treasury-coordinated bailout; one partner borrowed $24 million personally to invest more in the fund and ended up with a negative net worth
• Buffett: "How could 10 or 15 guys with an average IQ of maybe 170 get themselves into a position where they can lose all their money?"
• Thorp: LTCM's collapse traced directly to Merton and Scholes's intellectual rejection of the Kelly Criterion

Lessons across both careers

• Past learnings apply in unexpected future contexts: Shannon's wartime cryptography work directly shaped information theory; casino math directly shaped market strategy
• Give yourself time to think without external input — Shannon's best work came from extended solitary reflection
• Watch what people do, not what they say: revealed preference over stated preference
• Concentrated bets in your highest-conviction ideas beat diversification (Shannon, Carnegie, Buffett all converged on this)
• Never interrupt compounding — the RICO raid cost Thorp and his investors far more than any trading loss
• Start where you are: Thorp was 32 and broke before he ever thought seriously about finance