Beating Roulette: The First Wearable Computer
Thorp and Claude Shannon's roulette prediction device — arguably history's first wearable computer, decades ahead of its time.
Alongside the blackjack work, Thorp partnered with information theory pioneer Claude Shannon on an even more audacious project: a concealable computer, worn on the body and operated with hidden switches in a shoe, that used simple physics (timing a roulette wheel's and ball's rotation) to predict which roughly one-eighth of the wheel the ball was likely to land in — not a perfect prediction, but, like the blackjack system, a real, calculable statistical edge over the house, achieved through applied physics and timing rather than a memory-based mechanism the way card counting worked.
The book recounts testing the device in real casinos, including painful practical failures (wires shocking users, timing signals being thrown off by wheel wobble) alongside its real successes, and frames it as an early, little-known landmark in wearable computing — built and tested years before anything resembling a personal or wearable computer existed commercially — created purely to solve a specific, well-defined edge-finding problem, the same motivating spirit that drove all of Thorp's later work.
The roulette device is presented as revealing something more general about how Thorp approached every edge he pursued throughout his career, in and out of casinos: identify a system with real physical or mathematical structure (a biased wheel, a countable deck, a mispriced security), build a rigorous model of that structure, test it empirically with real stakes at small scale before committing serious capital, and only then scale up — the exact same sequence he later applied to warrant and options arbitrage and, still later, to the more complex strategies of his hedge funds.
- Thorp and Claude Shannon built a concealable, wearable computer to predict roulette outcomes using timing physics — an early landmark in wearable computing.
- Like blackjack, this exploited a real, calculable statistical edge, not a gambler's-fallacy-style pattern.
- The project reveals Thorp's general working method — model the structure, test small with real stakes, then scale — repeated throughout the rest of his career.