The Ritz laser scanner: £1.3 million in two nights, and why it was not a crime

Two nights in March

The Ritz Club is the casino under the Ritz hotel on Piccadilly, one of the more expensive rooms in London. In March 2004 a party of three, two men from Serbia and a woman from Hungary, played roulette there on two consecutive nights. On the first night they won about £100,000. On the second they won about £1.2 million, and the club, which had been watching, called the police.

The Metropolitan Police arrested the three and seized the cash and their equipment. The allegation, reported at the time, was that one of them carried a mobile phone containing a laser scanner. Pointed at the wheel in the first seconds after the dealer released the ball, it measured the speed of the ball and the speed of the wheel. A small computer, in the phone or on a person, turned those two numbers into a prediction: not a pocket, which is beyond any instrument, but a sector, a stretch of a few neighbouring numbers where the ball was most likely to come to rest. The prediction arrived before the dealer called no more bets, and the bets went on that sector.

The club had lost £1.3 million to it. What it had not lost was a single spin to anything that touched the table. The ball was the club’s ball, the wheel was the club’s wheel, the dealer spun both as usual, and every bet was placed inside the window the club itself allowed.

Why it works

Roulette is presented as pure chance, and for a bettor with nothing but eyes it is. But the ball, once released, is a physical object decelerating on a track under friction and air resistance, and the wheel is a rotor slowing at a rate that can be measured. Given the ball’s speed, the rotor’s speed and the geometry of the wheel, the point where the ball leaves the track and the sector of the rotor beneath it at that moment are not random; they are a calculation. The scatter after the ball hits the frets adds noise, which is why the target is a sector rather than a number, but the noise is not large enough to erase the edge.

None of this was new in 2004. In 1961 Edward Thorp and Claude Shannon built what is generally called the first wearable computer, a device the size of a cigarette packet that timed the wheel with a toe switch and signalled a predicted octant to an earpiece; Thorp later wrote that it gave an expected gain of about 44% on a bet. In the late 1970s a group of physicists in Santa Cruz, the Eudaemons, put the computer in a shoe and took it to Las Vegas, and their story became a book, The Eudaemonic Pie. What was new was a laser: a way to take the timing measurement without a human pressing a switch, quickly and precisely enough to make late betting comfortable.

The mathematics is the mirror image of the wheel-bias story. Jagger and García-Pelayo exploited a wheel that was slightly wrong and needed thousands of spins to find out how. The Ritz team exploited a wheel that was perfectly right, whose physics were so regular that a single spin could be predicted while it was still in progress. Bias is a defect. Prediction needs no defect at all.

Why it was not a crime

The three were arrested in March and bailed. In December 2004 the police announced that no charges would be brought and the money and equipment would be returned. The law in force was the Gaming Act 1845, and its provision on cheating was written for a world of loaded dice and marked cards: it punished winning by fraud or by an unlawful device used in the game itself. A laser that measured the game and a computer that reasoned about it did neither. The ball landed where the ball landed. The scanner had changed nothing but the players’ knowledge.

That is the same line the Spanish Supreme Court drew for García-Pelayo the same year: observation is not interference. It is also the line that separates the Ritz three from Phil Ivey, whose edge sorting case turned on the fact that he had asked for the cards to be turned, an act on the game rather than a measurement of it. English courts later found against Ivey on exactly that distinction. The Ritz team had asked the dealer for nothing.

The law did not stay where it was. The Gambling Act 2005, which replaced the Victorian statutes, contains a section on cheating that reaches actual or attempted deception and interference with the process by which gambling is conducted, and it is a criminal offence whether or not the cheat wins. Whether a laser scanner is “interference with the process” has not, as far as the public record shows, been tested in an English criminal court, and casinos have not needed to test it: they answer prediction with procedure, calling no more bets earlier in the spin, which removes the seconds the computer needs.

The wheel was honest, the ball was honest, and the house lost £1.3 million to two numbers it had let everyone see.the Ritz case in one sentence

What a hashed wheel gives away

It is worth being exact about why none of this transfers to an online wheel, because “online roulette cannot be predicted” is usually asserted rather than shown. On this site a roulette spin has no ball and no rotor. The pocket is derived by taking the round’s hash, reading its first eight bytes as a number between 0 and 1, multiplying by 37 and rounding down. There is no first second of the spin in which the outcome is in progress and measurable, because the outcome is not a process. It is a function of the committed server seed, your client seed and the round number, and it exists in full the moment the round exists.

What a hashed wheel gives away is different and, for a player, better. It gives away the commitment: before you bet you are shown the SHA-256 fingerprint of the server seed, and after the seed is retired you are shown the seed, so the one thing you can measure is whether the house kept its word. The verification walkthrough does that measurement on a live round. It will never tell you where the next ball lands, because nothing can. It will tell you, every time, that the last one landed where the seed said.

• Physical wheel: the outcome is a process. Measure it early enough and you can predict a sector. The house’s defence is procedure: close betting before the ball has told you anything.
• Hashed wheel: the outcome is a function. Nothing about it is in progress before the bet, so there is nothing to measure. The house’s obligation is disclosure: commit to the seed first, reveal it after.
• Both: the house edge is unchanged either way. A predicted sector on a fair wheel beats the 2.7% edge; a verified hash on a fair engine confirms it. Neither is a strategy for the other.