early or late they were in hitting the switch.
The results were ingenious, and perhaps doomed to fail. After much trial and error, Thorp and Shannon calculated a method to predict, with favorable odds, which octant of the roulette wheel the ball would tumble into. The wheel contained eight octants—six octants with five pockets each and two with four, making up the thirty-eight pockets on the wheel. If they could predict the octant, that tipped the odds sharply in their favor. If they bet on all four or five numbers in the predicted octant and their method proved accurate, winning would be guaranteed. It would be cheating, of course, and if they were caught, there was a predictably high chance that large, thick-necked casino bouncers with hairy knuckles would exact a price. But that was a concern for another day.
Thorp and Shannon designed a computer the size of a cigarette pack and embedded it in a pair of shoes. It had two switches: one switch turned on the computer, and the other timed the spinning of the rotor (one toe click when the wheel started and another when it made a single revolution). The computer calculated the results and transmitted which octant to bet on in eight tones to another person wearing a primitive sort of headphone in one ear. In all probability it was the world’s first wearable computer.
However, technical problems doomed the project. The headphone wires often broke. One time Thorp, who generally wore the headphone and placed the bets, noticed a woman staring at him with horror. He promptly headed for the bathroom. In a mirror he saw the speaker jutting from his ear like an alien insect.
Though Shannon didn’t lead Thorp to riches at the roulette wheel, the professor did make a key contribution to his younger colleague’s blackjack strategy. While Thorp had devised a winning approach to blackjack, a key unanswered question remained: how much should a bettor wager if he doesn’t want to risk financial ruin? Shannon told Thorp that the answer could be found in a 1956 paper by John Kelly Jr., a physics researcher at Bell Laboratories in Murray Hill, New Jersey. The paper described how much a gambler with inside information about the winner of a series of baseball games between two equally matched teams should wager if there is a certainamount of noise (and hence a possibility that the information could be faulty) in the channel communicating that information.
Thorp realized he could use Kelly’s betting system to optimally regulate how much he wagered on various scenarios in blackjack. In simplest terms, when his odds of winning rose, he tossed more chips on the table. When his odds got worse, he backed off.
A good way to size up Kelly’s system is by comparison with another gambling strategy: doubling down. Say you bet $10 on a hand of blackjack and you lose. If you bet $20 on the next hand and win, you’re up again. But you could lose that, of course. Bet $40, win, and you’re back ahead. Doubling down, also known as martingale betting, has been a time-honored strategy practiced by gambling legends such as Casanova. But there’s an obvious flaw in the strategy: gambler’s ruin. Eventually the martingale gambler will run out of money. The odds of this happening, if the gambler keeps playing, are 100 percent.
Kelly, however, limited the amount from a player’s billfold that could be placed on any bet. The only time a player would go all in would be when the odds of winning are 100 percent, a very rare event that almost never happens in a casino—although Thorp would discover such opportunities on Wall Street several years later.
The mathematics of Kelly told him exactly how much to add or subtract, based on the amount in his billfold, in order to achieve the maximum gains. The formula described, in the words of Kelly, how a gambler could “cause his money to grow exponentially,” while at the same time avoiding the curse of gambler’s ruin.
In January 1961, Thorp presented his