Wheel bias: the two men who beat roulette by writing numbers down, and why a court said it was legal

The arithmetic of a bias

A single-number bet on a European wheel pays 35 to 1, so a win returns 36 units on a 1-unit stake. With 37 pockets equally likely, the expected return is 36 ÷ 37, a shade over 97%, which is the house edge of 2.7%. The bet is unbeatable for exactly as long as the 37 pockets are equally likely. Change that assumption by a little and the sign flips.

Return = 36 × probability. A bias that moves one pocket from 1 in 37 to 1 in 30 is a difference of about six spins in a thousand, invisible to the eye and worth a fifth of every bet placed on it.

That last column is the reason the story exists. A worn fret, a slightly deeper pocket, a spindle a hair off true: any of them can tilt one number by a few tenths of a percent, and a few tenths of a percent on a 35-to-1 payout is a large edge. The catch is that the same smallness makes a bias almost impossible to see. Telling a 1-in-30 pocket from a 1-in-37 one with reasonable confidence takes several thousand recorded spins of that specific wheel, which at a real table is weeks of standing and writing.

Monte Carlo, 1873: the engineer with six clerks

Joseph Jagger was a Yorkshire textile engineer who understood, as a matter of trade, that machines are never perfectly true. He reasoned that a roulette wheel is a machine, hired six clerks, and had them record every result on the six wheels at the Monte Carlo casino for days. Five wheels showed nothing. One showed nine numbers turning up noticeably more often than their share.

He bet those nine numbers and, in the accounts that survive, won about two million francs over a few days, a fortune at the time. The casino responded the way casinos have ever since: it moved the wheels between tables overnight, which cost Jagger a day of losses until he noticed a scratch on his wheel and found it again, and then it had the wheel-maker fit movable frets that were rearranged daily. The bias vanished, Jagger stopped, and he went home with the money.

The dates and sums vary between tellings, and the music-hall song about the man who broke the bank at Monte Carlo, written in 1891, is usually attached to a different gambler. What does not vary is the method: no device, no accomplice, no touching the wheel. A notebook, arithmetic and patience.

Madrid, the 1990s: a family, a notebook, and a computer

A century later Gonzalo García-Pelayo, a Spanish record producer and film director who had run through several careers, did the same thing with better tools. Through the early 1990s he and members of his family recorded tens of thousands of spins at Casino Gran Madrid, entered them into a computer, and looked for pockets whose frequencies sat outside what chance would allow. The wheels were not all biased, but some were, and the biased ones were biased enough.

The family bet the favoured numbers and won steadily, reportedly the equivalent of well over a million euros in Madrid before taking the method to Las Vegas and elsewhere. The casino noticed, barred them, and went to law, arguing that the winnings were obtained by fraud. The case went up through the Spanish courts for years.

The player had done nothing but observe the wheel and bet on what he observed. If the wheel was imperfect, that was the casino’s failure, not the player’s deceit.the substance of the Spanish Supreme Court’s 2004 ruling in García-Pelayo’s favour, as reported

In 2004 Spain’s Supreme Court found for García-Pelayo. Exploiting a physical bias by watching and counting was legitimate; the remedy for a casino that did not like it was to maintain its wheels. He kept the money and his story became a feature film. Casinos everywhere now rotate wheels, rebalance them and run their own frequency logs, which is why the method is close to extinct in practice. The law, though, was settled in his favour.

Two courts, two answers, one line between them

Put the Madrid ruling beside the London one in Phil Ivey’s edge-sorting case and the whole law of advantage play fits on one line. Ivey also exploited a manufacturing flaw, a printing asymmetry on the backs of the cards, and also touched nothing. The English courts still called it cheating, unanimously, because he had caused the croupier to sort the deck for him under a false pretext. The game he played was not the game the casino thought it was dealing.

The flaw is not the deciding factor in any of the three. The player’s conduct is.

Observation is on one side of the line and manipulation on the other. Both Jagger and García-Pelayo took the game exactly as the house offered it and were simply better informed about it than the house was. Ivey changed the game. Courts in two countries, sixteen years apart, drew the line in the same place.

No wheel to wear

The method cannot be tried here, and the reason is worth stating precisely rather than as a slogan. There is no wheel. The pocket for a spin is a number derived from the committed server seed, your client seed and the round number, and the derivation treats all 37 pockets identically.

A uniform number multiplied by 37 and rounded down gives each pocket a probability of exactly 1 in 37, to the resolution of the hash. Nothing wears, nothing tilts, and there is no daily maintenance to skip. What a modern García-Pelayo could do instead is the check the scheme is built for: after a server seed is revealed, every spin played under it can be recomputed, and a long enough log of revealed spins can be tested for the uniformity the code promises. Four ways to prove a random number puts that check in context, and the verification walkthrough does it on a live round.

It is a smaller freedom than the one the Spanish court protected, and a more useful one. Jagger and García-Pelayo could find out that a wheel was unfair only by betting on it for weeks. Here the question is answerable before the first bet, and the answer, for every pocket, is 1 in 37.