The Roulette and Market Edge

Wearable Computers in Las Vegas and Constant-Rebalanced Portfolios With Ed Thorp

Down in the basement workshop of a lakeside house in Winchester, Massachusetts, in early 1961 stood a regulation casino roulette wheel purchased for fifteen hundred dollars. Late in 1960, a young Massachusetts Institute of Technology mathematics instructor named Edward O. Thorp had knocked on Claude Shannon's office door in Cambridge. Thorp had already discovered a card-counting system that beat casino blackjack by tracking the depletion of small cards against tens and aces. Now he wanted to show Shannon a second project. Thorp believed a player could predict roulette spins by treating the spinning rotor and ivory ball as a deterministic Newtonian clockwork.

Shannon invited Thorp out to his three-story Victorian house—which he and Betty called Entropy House. Together the two mathematicians spent months filming roulette spins with high-speed strobe cameras. In standard American thirty-eight-pocket roulette (with 0 and 00), the casino enjoys a house edge of $2/38 \approx 5.26\%$ so long as every pocket is equally likely. However, a croupier invites bets while the wheel rotor spins in one direction and the ball whirls around the tilted wooden track in the opposite direction.

Shannon and Thorp found that once a roulette wheel has a slight tilt of even 0.2 degrees, a predictable pattern emerges. The ball always drops off the upper track at nearly the same downhill spot when it slows to a critical velocity. Measuring the angular velocity and deceleration of the ball and the rotor across two revolutions predicts the landing octant of the wheel with an expected player advantage of $+44\%$. To exploit that edge inside a casino, they soldered together the world's first wearable computer between November 1960 and June 1961. The device was a twelve-transistor analog timing circuit the size of a cigarette pack, worn at the waist and controlled by hidden microswitches inside the soles of their shoes.

In the summer of 1961, Claude and Betty Shannon joined Ed and Vivian Thorp in Las Vegas. Standing at the roulette rail, Shannon tapped his big toe inside his shoe as the rotor zero mark and the ball flashed past a reference diamond. The pocket computer integrated the two angular rates and played a musical pitch into a tiny loudspeaker inside Thorp's ear canal. Each tone signaled which of the eight wheel octants to bet on just before the croupier called no more bets.

That Las Vegas expedition also connected directly to the problem of optimal capital allocation under uncertainty, pioneered at Bell Labs in 1956 by Shannon's colleague John L. Kelly Jr. While Kelly showed what fraction of a bankroll to wager when a gambler holds a positive statistical edge, Shannon surprised an MIT lecture hall in the mid-1960s with a counterintuitive extension. He demonstrated how an investor can extract guaranteed compound growth from a volatile stock whose average drift is zero—a puzzle later dubbed Shannon's Demon.

Imagine a volatile stock whose price either doubles ($\times 2$) or falls in half ($\times 1/2$) every day with equal $50\%$ probability. A buy-and-hold investor who puts $100\%$ of their wealth into that stock goes nowhere over time. One doubling followed by one halving multiplies total wealth by $2 \times \frac{1}{2} = 1.00$. Now consider Shannon's rule: keep exactly $50\%$ of your wealth in cash and $50\%$ in the stock, rebalancing back to $50/50$ at the close of every single day.

When the stock doubles, the cash half stays at 1 and the stock half grows to 2. Total portfolio wealth is multiplied by $\frac{1}{2}(1) + \frac{1}{2}(2) = \frac{3}{2} = 1.50$. When the stock halves, the cash half stays at 1 and the stock half drops to $1/2$. Total wealth is multiplied by $\frac{1}{2}(1) + \frac{1}{2}\left(\frac{1}{2}\right) = \frac{3}{4} = 0.75$. Thus the expected logarithmic growth rate $g$ per step is strictly positive:

Because rebalancing forces the investor to sell shares after every doubling and buy cheap shares after every halving, volatility itself is harvested into exponential growth. By rebalancing fifty percent cash against fifty percent of a volatile zero-drift stock after every price swing, Shannon's mathematical portfolio compounded wealth by twelve and a half percent across every up-and-down cycle while a buy-and-hold investor broke exactly even.