Cheating at Roulette

[Robin]

A while ago, on a documentary about gambling, I heard of a group of university students who built a computer to cheat at roulette. The idea was that there is a switch in your shoe and you click it every time the ball passes a certain number, then the computer calculates the angular velocity of the ball and gives feedback about which number to bet on (as a series of clicks against the skin).

Now this whole thing appears to me to be absolutely impossible. Sure you could probably calculate the angular velocity that way, but the way the ball bounces will depend on precisely how it hits those bevelled edges.

I have heard it referred to in two documentaries and each time completely uncritically.

Has anybody heard of this machine and whether it is real or not?

 

[whitefork]

Read all about it
http://www.amazon.com/gp/product/05...f=pd_bbs_1/103-1578809-8195828?_encoding=UTF8

 

[fabian_lidman]

I recall seeing this in an episode of C.S.I.

 

[lenny]

A while ago, on a documentary about gambling, I heard of a group of university students who built a computer to cheat at roulette.

it was not technically cheating, at the time...

Now this whole thing appears to me to be absolutely impossible. Sure you could probably calculate the angular velocity that way, but the way the ball bounces will depend on precisely how it hits those bevelled edges.

they did not attempt to use equations of motion; they used a data based approach.

when predicting real systems when you can get a lot of data (not just lots of data points, but long in terms of Poincare returns), data based models almost always win, especially those where the best models are chaotic.

the book is by Thomas Bass, it has different titles on each side of the Atlantic; the one i read was "The Newtonian Casino".

 

[Ersby]

This was written up in a book called "The Newtonian Casino" which, unfortunately, I can't find at the moment so I'm at a loss to tell you any details about the people involved.

As I recall, the behaviour of the ball hitting those studs was taken into account, and the computer could tell the player which 90 degree space on the wheel the ball was likely to land in.

I'll have another look and see what I can find.

(It's a good book, btw)

ETA: me so slow.

 

[a_unique_person]

Read all about it
http://www.amazon.com/gp/product/0595142362/sr=8-1/qid=1148217818/ref=pd_bbs_1/103-1578809-8195828?%5Fencoding=UTF8

Sort of like Leonard Nimoy singing "Proud Mary."

I won't be reading that...

 

[andyandy]

I'm sure people have tried to use computers to beat the roulette odds......all it would need would be to tilt the odds in your favour rather than the house's favour to be profitable in the long run....

there seems to be some annecdotal evidence that people are using devices - although whether or not they change your odds remains debatable......

Earlier this year a Hungarian gambler, Laszlo Kovacs, installed a small computer in his shoe, activated by tapping. Using factors such as the velocity of the roulette wheel when it's spinning, the computer calculated what number would show up next.

However, Kovacs ran out of luck and was ultimately arrested by Australian police for using roulette cheats. According to authorities' estimates, Kovacs collected approximately $200,000 in various casinos.

In March of this year, two Serbian men and a Hungarian woman were accused of collecting about $2.4 million in winning at roulette, with the help of a laser scanner in a cellular phone.

The police now needed to figure out if the scanner was actually capable of predicting in which pocket the would fall. Experts in the gaming industry are not easily convinced that the laser device is fit for this mission, or, in fact, that any other apparatus can cope with the task. According to UK gaming law, the ball must circle the roulette wheel no less than three times. After three spins, the dealer announces "no more bets." The laser device must be quick enough to calculate the outcome in merely a few seconds. Although at present, there are no specific rules about cheating in casinos, there are various government proposals that do address the prohibition of outside influences on casino gaming.

http://www.roulette2002.com/cheats2.php

 

[Robin]

it was not technically cheating, at the time...

They seemed to go to a lot of trouble to hide the equipment - input/output devices in their shoes, even suffering burns from a leaky battery secured around the waste..

they did not attempt to use equations of motion; they used a data based approach.

Both documentaries suggested they were using the equations of motion, which is why I thought that they did.

when predicting real systems when you can get a lot of data (not just lots of data points, but long in terms of Poincare returns), data based models almost always win, especially those where the best models are chaotic.

the book is by Thomas Bass, it has different titles on each side of the Atlantic; the one i read was "The Newtonian Casino".

Thanks, I will see if I can track down the book (preferably in a library, I am still skeptical and wouldn't want to put money in his hands if it is just a hoax).

 

[Hawk one]

As I recall, the behaviour of the ball hitting those studs was taken into account, and the computer could tell the player which 90 degree space on the wheel the ball was likely to land in.

That was basically my thoughts. That this method would only be able to give a certain likelyhood of the area that the ball would end up in; not claiming that it would predict the exact number.

And if that's possible - which I cannot say myself, of course, as I've never heard about this before today - then it will be more than enough give you a huge advantage, since the table will be reduced from 37 equally possible targets, to 9 targets that are more likely than the rest.

 

[Robin]

Earlier this year a Hungarian gambler, Laszlo Kovacs, installed a small computer in his shoe, activated by tapping. Using factors such as the velocity of the roulette wheel when it's spinning, the computer calculated what number would show up next.

This sounds like the set up that I heard about. Sounds to me like you would have to have supernaturally quick reflexes to be able to use this, even if it did work.

 

[Robin]

As I recall, the behaviour of the ball hitting those studs was taken into account, and the computer could tell the player which 90 degree space on the wheel the ball was likely to land in.

This is what seems unlikely to me. Even if you had a really precise observation of the ball's speed and position you probably couldn't get a meaningful model of it's behaviour hitting a bevel.

 

[Robin]

That was basically my thoughts. That this method would only be able to give a certain likelyhood of the area that the ball would end up in; not claiming that it would predict the exact number.

And if that's possible - which I cannot say myself, of course, as I've never heard about this before today - then it will be more than enough give you a huge advantage, since the table will be reduced from 37 equally possible targets, to 9 targets that are more likely than the rest.

I don't know that this would be a huge advantage. It sounds like you are still losing more than you win.

 

[lenny]

I'm sure people have tried to use computers to beat the roulette odds......all it would need would be to tilt the odds in your favour rather than the house's favour to be profitable in the long run....

this is not difficult to do: if the pay off is 36 for 1 and the (naive) odds or 37 or 38 to 1, depending on where you play; then if you can get the quarter of the wheel correct half the time...

there seems to be some annecdotal evidence that people are using devices - although whether or not they change your odds remains debatable......

it is well documented and easily demonstrated.

afterall, the original (public) story was 20-ish years ago... should be easy now with a PC

i believe the laws in Nevada were changed after they went public...

 

[Hawk one]

This is what seems unlikely to me. Even if you had a really precise observation of the ball's speed and position you probably couldn't get a meaningful model of it's behaviour hitting a bevel.

Actually, I think the idea is that the hitting angle on the bevel is more or less accounted for. If I'm understanding this correctly, it's not about guaranteeing that the ball will fall into a specific 90 degree area, it's about being able to give a certain increased likelihood of where the ball will begin to fall down -at first-.

After that, one can make a rather crude statistic that shows where the ball has a tendency to end up after it makes the first impact. You don't really need to bother with hitting angles here, just evidence on the likes of "if the ball first hit Area X at first (which is what this counter is really predicting), then it lands in Area X 23% of the times, Area Y 27% of the time, Area Z 26% of the time, and Area A 24% of the time" (Pure abstract numbers here, of course). If you just repeat the motion enough times - say a few thousand repetitions - then that figure should be what's accounting for the hit angle. And all it needs to show is that if the ball hits Area X in the first impact, then it will be more or less likely that it will hit Area X, Y, and so on. And if there are in fact differences going on here (instead of 25% on all areas), then I'd think we have something going here. Probably not by much, but if you can manage to calculate your way into saying that the probability of hitting Area X from 25% to, say, 27%, that should be enough to move the odds to your favour...

Of course, I may still have misunderstood the entire idea here. My layman's brain just seems to think that there are (at least) two formulas needed to make this work, and this is but my suggestion.

 

[Terry]

Their strategy was to predict which quarter of the wheel the ball would land in, and to cover that quarter with bets. They practiced a long time to get the bets down in a timely manner, since of course the layout of the table doesn't put numbers which are adjacent on the wheel close to each other. I highly recommend the book (They Eudaemonic Pie), a very entertaining read.

 

[Robin]

it is well documented and easily demonstrated.

Respectively where and how? There has been one book mentioned, but I don't know how much detail it goes into. I have been trying to think of some way to easily demonstrate this, but I can't. Has the original code ever been published?

It seemsthat if I could rent or borrow a regulation size roulette wheel one could simulate this with a PC and demonstrate how much it would change the odds.

It would also help to demonstrate that it can be acheived with human reaction times in the limited amount of time between the ball going into motion and the bets being stopped.

 

[lenny]

I highly recommend the book (They Eudaemonic Pie)

methinks it "The" not "They"; published as "The Newtonian Casino" in Europe. the eudaemonic pie idea started before, and lived on well after the roulette attempt.

 

[Bob Klase]

It's been a long time since I've been at a roulette wheel, so I may be wrong. But don't they stop the betting as soon as the ball is dropped or shortly after that? How could you let it go around enough times to click the data with your toe, and then get any bets down before the betting was stopped?

 

[Robin]

It's been a long time since I've been at a roulette wheel, so I may be wrong. But don't they stop the betting as soon as the ball is dropped or shortly after that? How could you let it go around enough times to click the data with your toe, and then get any bets down before the betting was stopped?

That was one of my questions. Someone said earlier that UK gaming laws mandated 3 complete circuits before betting is stopped. I am not sure what the rules were in Nevada in the 70's but I have stood in a casino and tried to imagine the timing (click, click, feedback, make decision, bet) and it seemed impossible to me. It all happens too quickly.

I gotta say my BS detector is buzzing at this one.

 

[lenny]

Respectively where and how? There has been one book mentioned, but I don't know how much detail it goes into. I have been trying to think of some way to easily demonstrate this, but I can't. Has the original code ever been published?

i expect there is enough info in the book to guess the algorithm; there is also a PRL in 1987 (Farmer and Sidoriwich) which gives a much more complicated data based modelling approach (and sparked off a great deal of interest in forecasting chaotic systems). if you are serious about constructing a demo, i could happily supply code (C) and info on constructing a learning set, in exchange for access to the data for teaching &c. (send a pm). i expect a simple analogue based approach would suffice (a "whether model"), but as you say it would be interesting to see just how skillful one could get the forecast probabilities.

It seemsthat if I could rent or borrow a regulation size roulette wheel one could simulate this with a PC and demonstrate how much it would change the odds.

if you can get the I/O to work smoothly, i expect that it is straight forward; i'd naively give 2:1 odds the rate limiting step is placing the bets. do you see an approach better than quarters of the wheel?
would you be happy with getting the quarter correctly 50% of the time? 30%?
would you insist on betting every time, or only on a strong signal?

and of course, if you wanted only proof of concept, you could easily take loads of data without actually placing bets, including the time the betting was closed, and then factor in the processing time, how long it took to place the bets, ...

you could also use film of past games in class, and have students play in real time, each using their keyboard... might be a good way to show people there can be nontrivial profit in NOT using matlab exclusively...