Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Math Idle

How a Professor Beat Roulette, Crediting a Non-Existent Supercomputer (thehustle.co) 156

I loved this story. The Hustle remembers how in 1964 a world-renowned medical professor found a way to beat roulette wheels, kicking off a five-year winning streak in which he amassed $1,250,000 ($8,000,000 today). He noticed that at the end of each night, casinos would replace cards and dice with fresh sets -- but the expensive roulette wheels went untouched and often stayed in service for decades before being replaced. Like any other machine, these wheels acquired wear and tear. Jarecki began to suspect that tiny defects -- chips, dents, scratches, unlevel surfaces -- might cause certain wheels to land on certain numbers more frequently than randomocity prescribed. The doctor spent weekends commuting between the operating table and the roulette table, manually recording thousands upon thousands of spins, and analyzing the data for statistical abnormalities. "I [experimented] until I had a rough outline of a system based on the previous winning numbers," he told the Sydney Morning Herald in 1969. "If numbers 1, 2, and 3 won the last 3 rounds, [I could determine] what was most likely to win the next 3...."

With his wife, Carol, he scouted dozens of wheels at casinos around Europe, from Monte Carlo (Monaco), to Divonne-les-Bains (France), to Baden-Baden (Germany). The pair recruited a team of 8 "clockers" who posted up at these venues, sometimes recording as many as 20,000 spins over a month-long period. Then, in 1964, he made his first strike. After establishing which wheels were biased, he secured a £25,000 loan from a Swiss financier and spent 6 months candidly exacting his strategy. By the end of the run, he'd netted £625,000 (roughly $6,700,000 today).

Jarecki's victories made headlines in newspapers all over the world, from Kansas to Australia. Everyone wanted his "secret" -- but he knew that if he wanted to replicate the feat, he'd have to conceal his true methodology. So, he concocted a "fanciful tale" for the press: He tallied roulette outcomes daily, then fed the information into an Atlas supercomputer, which told him which numbers to pick. At the time, wrote gambling historian, Russell Barnhart, in Beating the Wheel, "Computers were looked upon as creatures from outer space... Few persons, including casino managers, were vocationally qualified to distinguish myth from reality." Hiding behind this technological ruse, Jarecki continued to keep tabs on biased tables -- and prepare for his next big move...

In the decades following Jarecki's dominance, casinos invested heavily in monitoring their roulette tables for defects and building wheels less prone to bias. Today, most wheels have gone digital, run by algorithms programmed to favor the house.

This discussion has been archived. No new comments can be posted.

How a Professor Beat Roulette, Crediting a Non-Existent Supercomputer

Comments Filter:
  • Today, most wheels have gone digital, run by algorithms programmed to favor the house.

    Is that even possible? Doesn't roulette give you known odds of winning based on where you place your bet? Does the table know where bets are placed and change the number based on that? Seems unfair.

    • by Patent Lover ( 779809 ) on Sunday May 26, 2019 @05:19PM (#58658892)
      They mean instead of a mechanical wheel, it's a digital wheel which is not prone to non-randomness due to physical defects in the wheel.
    • In all the years I've visited casinos (going back as far as the con artist's failed ones in Atlantic City), I have never seen a digital roulette wheel. Without exception, every single one has the croupier spin the wheel, wait a few moments, then spin the ball in the opposite direction to the spin of the wheel.

      People keep using the word digital. I do not think it means what they think it means.

      • by Anonymous Coward

        Digital, as in housing pressure sensitive pads, electronically controlled balanced center of gravity.

      • by guruevi ( 827432 )

        Digital as in tracking video and varying the speed of slowdown. An algorithm can predict where the ball will land and calculate how much it needs to slow down to make it land in a more favored position. As with slot machines, the house does have legal restrictions on how much it can favor itself.

        • by lgw ( 121541 )

          Which would be illegal, at least in the US. The house advantage is built into the green slots, and the house can't be deceptive about the odds. They can have a machine-controlled roulette wheel, but the odds are constant.

          • by fred911 ( 83970 )

            "The house advantage is built into the green slots"
            Only on outside wagers. On inside wagers the house pays 35 to 1 on 36 possibilities and some wheels also include 00 providing the house with an even better advantage.

      • People keep using the word digital. I do not think it means what they think it means.

        Did you say the croupier uses his fingers?

    • by Solandri ( 704621 ) on Sunday May 26, 2019 @05:22PM (#58658910)
      TFA is just trying to make digital gambling sound more nefarious. Physical roulette wheels are already biased to favor the house [roulettesites.org]. e.g. You can bet on a red or black tiles at even odds. But there are one or two green tiles which result in the house auto-winning, which skews your even odds into losing odds. A perfect digital simulation of a physical roulette wheel thus has to be programmed to favor the house, because the physical roulette wheel is designed to favor the house.
      • by mark-t ( 151149 )
        Of course, but is possible to make a digital simulation of a roulette wheel that favors the house even more than a physical one does by taking into consideration all bets and may elect, in any given spin, to make the virtual "ball" land on a result that minimizes the payout, resulting in odds that favor the house even more.
        • by gweihir ( 88907 )

          And that would be a criminal act. Hence these machines have to be independently reviewed and certified.

    • by Namarrgon ( 105036 ) on Sunday May 26, 2019 @05:33PM (#58658942) Homepage

      I used to write software for slot machines back in the 90s, replacing the old mechanical reels with digital equivalents. Their payback was statistically determined by the fixed layout of the symbols on their "reels", just like the old machines, only driven by an RNG with well-known randomness instead of unreliable mechanics. They were scrupulously fair (by law), so long as you understood that on average you would only get back 86% of what you put in (in fact buyers could choose machines with specific payout percentages, to better suit their venue's usage and customers etc).

      I imagine modern roulette wheels are similar. The odds are determined by the wheel layout as before, are still tilted towards the house, but are now strictly identical from roll to roll.

      • by fatwilbur ( 1098563 ) on Sunday May 26, 2019 @06:57PM (#58659298)
        You are spot on correct. I have worked on gambling machines as well and they are 100% random, however through basic statistics you can determine almost exactly how much will be paid out - over the long term. I still play them, I think it's fun and know it's purely for entertainment. It is possible to win, you just need to be able to walk away once you're up, realizing that the longer you play, you will eventually revert to the mean (losing).

        Interesting side note, is just how prevalent the Gambler's fallacy [wikipedia.org] can be seen among most everyone else I meet playing. People are convinced they know when a machine is about to pay out, when it's "hot" or "cold", when it's been "primed" etc. All false, but I do understand it and even suffer from it myself - there are certain days I feel lucky, even if I understand it's just an irrational conclusion of my brain.

        If you want a good point by point breakdown of how they generally work, the governing body of gaming in my province published a good summary [aglc.ca]:

        -Vlt machines in Alberta are programmed to pay out a cumulative average of 92 per cent of all money wagered over all of the machines in the casino network; the payout rate of 92 per cent is averaged over millions and millions of plays, not one playing session
        -The programmed machine payout rates vary based on line counts, denomination, maximum bet and various other variables; for example, some machines may be programmed at 92.00 per cent or 91.89 per cent
        -It’s important to note that vlts operate randomly at all times, no matter how many wins or losses have occurred in the past; altering play patterns, time of day, or position of the machine within a video lottery retail location or the result of your last game has no bearing on the result of your next game
        -Each spin on a vlt is random and is independent of the amount wagered; this means that the machine randomly picks the outcome and then evaluates it against the wager to determine the prize amount
        -Every time the “Spin” button is pressed on a vlt machine, including progressives, the outcome is determined by a computer chip called a Random Number Generator (RNG) located inside each machine; a new independent random draw is made each time the “Spin” button is pressed then computer program within the machine uses sophisticated random mathematical processes to determine the outcome of each game
        -The random draw cannot be influenced by anything other than the RNG and there are not patterns or clues to indicate when a win will happen; randomness means that all future outcomes are unpredictable
        -No matter the type of game you are playing, the outcome of every bet is ultimately determined by random numbers; the game will choose one random number for each reel, map that number onto a position on the reel, stop the reel in the appointed place and score whatever the outcome is; the outcome is randomly determined at moment you press the button; your odds are the same for every spin on a given machine
        -The odds of winning smaller prizes are much greater than the odds of winning the top prize; the laws of probability will allow players to win, lose or leave and others to walk away even
        • by Namarrgon ( 105036 ) on Sunday May 26, 2019 @09:47PM (#58660014) Homepage

          Indeed. I found it interesting that the outcome is all determined at the moment the user presses Spin - all completely random, but the results of the spin, any optional double-ups etc are all set at that moment, and the spinning of the reels merely traditional, to build anticipation. In fact they balance that spin time against throughput, because they want to enable as many individual spins per minute as they can get away with.

          In our area, payback factors generally ranged from 0.83 to 0.88, but could be as high as 0.92 in high-traffic venues - again, balancing their net take against enticing users to keep going with more payouts. I've heard that slots in Vegas can get right up to 0.97, giving away lots of wins to attract more punters, and making it up on sheer volume.

          • by hawk ( 1151 )

            > I've heard that slots in Vegas can get right up to 0.97, giving away lots of wins to attract more punters, and making it up on sheer volume.

            Some actually go as high as a *theoretical* 102%--for theoretically "perfect play".

            There generally aren't more than a handful of these anywhere.

            hawk

      • My understanding is that slot machine software actively varies the odds to game the gamer.

        Sure, they might provide an 86% return *on average*. But not on each bet. And the size also varies.

        So when a new player starts, it is good to give them a few small wins early on. Then, after a while, flip the odds to only give them a very occasional large win. If they begin to slow down, then give them a few smaller wins.

        Etc.

        • That would definitely be illegal under the gambling regulations I worked with.

        • by gweihir ( 88907 )

          No. Easy to spot and would lose the ones doing it not only their license, but would likely land them in prison. These machines are highly regulated.

        • Hmm, I remember reading an article a few years ago about this- that slot machines were programmed to play games with the odds between plays and that when people took it up with the gaming commission they said as long as the average odds were correct all was fine. Unfortunately I can't remember where or what gaming commission it was nor can I find the article so...
      • by Trogre ( 513942 )

        I probably put your kids through college back in the 90s.

    • by guruevi ( 827432 )

      Pretty much, these days a camera tracks bets and the speed of the ball and it slows down the wheel to make the ball land in an area that will maximize the profits (although many places do have regulations on how much the house may favor itself). Back in the day they likewise used brakes or magnets on the tables but it took more skill from the crew to balance the odds in favor of the house.

      • These days a camera tracks bets and the speed of the ball and it slows down the wheel to make the ball land in an area that will maximize the profits

        Citation? Or I call shenanigans on your opinion on how casinos cheat the players.

        They already have a built-in advantage. Why risk losing their license for a few dollars more?

      • by gweihir ( 88907 )

        No. Not if it is a licensed casino. In an illegal one, sure, expect all kinds of criminal fraud. A licensed one is hugely profitable when respecting the respective laws and they will not lose their license and worse by doing such things as you describe.

      • Have you any evidence of this? It's _possible_, but puts the house's normally consistent chance of winning at risk if they are caught, or if a cheater can suborn the system. If a roulette player can bet _against the other players_, they could gain a noticeable advantage over the house.

    • I suppose that part of the article was just made up.
    • by jrumney ( 197329 )
      Every form of gambling favors the house. Why would casinos exist if there wasn't some a pretty much guaranteed profit stream for the owners? In the case of roulette, they odds are made to look like they are even, but only if you overlook the green 0 (and on US wheels, the 00 opposite it) in your calculations.
      • Actually the form of Gambling where Trump is involved favors the house going bankrupt.
      • by gweihir ( 88907 )

        So the casino greed is worse in the US? No surprise there.

        • by jrumney ( 197329 )
          It's largely historical, but yes, initially US based casinos gave a 5.6% margin to the house for roulette by putting an extra 0 in the wheel, where in Europe it was always 2.8%. US casinos started moving to European wheels to gain customers many years ago though.
    • by gweihir ( 88907 )

      If you have a uniform distribution, the house is favored. The "0" is full win for the house and it comes up about 2% of the time. Simple, obvious statistics.

      Assuring uniform distribution is easy when done electronically. That is if you know what you are doing, but Casinos have enough money to hire actual experts instead of the typical low-skill, low-insight code monkeys used to often these days.

    • Not plausible at all ...

    • They could cheat - but random with a take-out may be the best way to extract money from customers long term anyway.

    • "Is roulette a game of luck?"
      "Not the way we play it here, no".

  • Hey (Score:5, Insightful)

    by phantomfive ( 622387 ) on Sunday May 26, 2019 @05:10PM (#58658860) Journal
    Hooray, Gambling! [smbc-comics.com]

    I have a friend who wrote a computer program to figure out how to beat a particular game when he was visiting some casino resort. He went all-in on it and had won $400-$500 in the positive on the trend using his strategy. During a break, he went back to the hotel room and noticed a bug in his program. Turns out his strategy was losing, not winning after all. He had just gotten lucky.
    • by gweihir ( 88907 )

      Typical observation when you do small samples. This is called a "statistical anomaly". Amateurs make the effect worse by selective memory. They remember far better when they won than when they lost. This fundamental incapability many people have to estimate probabilities realistically is what the whole gambling industry is build on.

  • by gnasher719 ( 869701 ) on Sunday May 26, 2019 @05:14PM (#58658874)
    Having a digitally simulated roulette wheel that isn't random but favors the house would be fraud and get the police involved very quickly.

    A game casino makes its money, slowly but steady, by having 37 (38 in the USA) numbers and rewarding wins with a factor 36 only. If you cover every number plus the zero (and the double zero in the USA), it costs you $37 ($38 in the USA) and you win $36 - guaranteed. That's what makes the casino a winner. Cheating on top of that would be idiotic, because as soon as it is discovered, you're out of business.
    • All roulette wheels favor the house due to paying out 36-1 on a number when the odds are 1 in 38, due to the 0 and 00 numbers on the wheel.
    • That's right. I'm quite sure the odds haven't changed - they don't need to, as you say - but the casino owner can now be assured that the likelihood of their digital wheel picking any specific number is now precisely 1 in 37 (or 38), as guaranteed by software instead of imperfect mechanics. In light of TFA, you could argue this actually favours the house more by eliminating these exploits.

    • Having a digitally simulated roulette wheel that isn't random but favors the house would be fraud and get the police involved very quickly. A game casino makes its money, slowly but steady, by having 37 (38 in the USA) numbers and rewarding wins with a factor 36 only. If you cover every number plus the zero (and the double zero in the USA), it costs you $37 ($38 in the USA) and you win $36 - guaranteed. That's what makes the casino a winner. Cheating on top of that would be idiotic, because as soon as it is discovered, you're out of business.

      Seems to me that the casinos can cheat with impunity, they just call it a malfunction. Not making that up, citations below: http://777click.com/slot-machi... [777click.com] https://www.techdirt.com/artic... [techdirt.com] https://www.insideedition.com/... [insideedition.com] https://abc7.com/news/woman-th... [abc7.com] here's a fun one where Indians get to play on tribal sovereignty to scam people: https://www.azcentral.com/stor... [azcentral.com]

      So many to choose from, it's a wonder that anyone gambles since casinos can decide to pay or not with no effective recourse. You lose -

      • by AmiMoJo ( 196126 )

        What boggles the mind is that the developers don't simply add a line that says

        if (jackpot > max_jackpot) jackpot = max_jackpot;

        You would think that would be one of the basic sanity checks they do.

        • What boggles the mind is that the developers don't simply add a line that says

          if (jackpot > max_jackpot) jackpot = max_jackpot;

          You would think that would be one of the basic sanity checks they do.

          Why would they do that when instead it becomes:

          if (jackpot > max_jackpot) jackpot = free_buffet;

          Call me cynical but it sure saves the casino money that they handle it the way they currently do. Of course it isn't fair but since when was gambling about fairness?

    • by mark-t ( 151149 )

      Having a digitally simulated roulette wheel that isn't random but favors the house would be fraud and get the police involved very quickly.

      How would the police even find out that a simulated wheel wasn't truly random?

    • Why would it be fraud to have the wheel come to a controlled stop instead of coming to a natural stop? The wheels in slot machines don't naturally come to a stop either. Everyone knows casinos are rigged.
      • by gweihir ( 88907 )

        What "everyone" knows is (as usual) wrong. Casinos are not "rigged". They win by design in the long run and the rates they win at are controlled by law.

    • by gweihir ( 88907 )

      You can digitally simulate a completely fair and legal and _random_ roulette wheel, no problem. What are you talking about?

    • by AmiMoJo ( 196126 )

      Having a digitally simulated roulette wheel that isn't random but favors the house would be fraud and get the police involved very quickly.

      That's how normal roulette wheels work. They have an equal number of red and black numbers, plus one green that tips the odds in the casino's favour.

      The odds are actually regulated by law in most places, and the legal authorities like software because it can be audited and tested to prove that the casino isn't cheating by exceeding the legal maximum bias in their favour.

    • by hawk ( 1151 )

      >Having a digitally simulated roulette wheel that isn't random but favors the house would be fraud and get the police involved very quickly.

      In Nevada, it would never even. pass gaming approval.

      Each and every ROM needs to go through thorough testing (although about 20 years ago, a video poker ROM with a cheat mode managed to get through. If a certain sequence of keys were hit, it went to cheat mode, and had something like a 90% chance of not dealing the last card of a royal when it came up. Unplug, and

  • by goombah99 ( 560566 ) on Sunday May 26, 2019 @05:18PM (#58658888)

    THe statement
    "If numbers 1, 2, and 3 won the last 3 rounds, [I could determine] what was most likely to win the next 3...."

    is thought provoking. Naively the zeroth order approximation would be a simple bias for certain numbers over others. The next order bias might be how the dealer spins the wheel. Why would this matter to the last number? it's unclear but perhaps in retreiving the ball from the prior spin the wheel is moved so the ball is in easy reach. This then lightly biases the starting position of the wheel and the roll of the ball if the dealer is reproducible.

    But how do we go back 2 more steps in the markov chain? Well perhaps the dealer isn't the same every night so the spins for any given dealer might be reproducible but the dealers vary. Going back two steps may tell you a bit about the dealers english on the spin.

    This however seems like very tenuous reasoning. Maybe there is some forehead slap someone else can give me for why the last 3 spins would matter.

    As for the odds, if were betting on straight up numbers then on a fair table the odds of winning are 1/38 and the payoff I beleive is 36:1. So one has a 36/38 loss rate on a fair table. So the bias has to be larger than 36/38. That seems like a lot of bias.

    Moreover this distribution is extremely skew. It's very easy to go a long time without a win. So there's a distinct chance of going bankrupt before the house goes bankrupt even when the odds are slightly in your favor.

    Still the odds don't have to be too much in your favor for this to add up. just a 1% edge on 1000 spins gives you a 20,000 fold gain.

    But I still can't figure out how the prior spins matter much.

    • But I still can't figure out how the prior spins matter much.

      Let's say you figure out the ball travels on average 7.29 rotations of the wheel, then you could use that to predict the next number. But by having the last three numbers, you could figure out whether the croupier increases or decreases his speed, and predict better.

      I'd assume that the force the croupier uses, and the delay that he uses, would change over time in a slightly predictable way.

      • by kaybee ( 101750 )

        I don't think this explains how the previous number(s) matter. When I have seen roulette, the dealer spins the wheel and then adds the ball. At this point the wheel is spinning fast enough I don't think anybody could time the placing of the ball. It seems to me that every spin is completely independent of all previous spins.

        The only possible pattern I could imagine would be just by looking at average numbers. If a wheel has imperfections perhaps it is possible that a given number or set of numbers happen mo

        • by hawk ( 1151 )

          >I don't think this explains how the previous number(s) matter.

          It likely determines the starting position of the wheel before spinning.

          If the dealer tosses the ball in a similar manner most of the time, that one defeated random number *might* give a tiny edge--or, it might be overwhelmed if the throw is "random enough" . . .

          I'm certainly skeptical that it can provide the nearly 3% edge (single 0) or nearly 6% (00) edge needed to compensate for the house edge.

          I considered serious sports betting years ago,

    • by burningcpu ( 1234256 ) on Sunday May 26, 2019 @08:31PM (#58659732)
      I'm guessing that the 'fail points,' or conditions that increase the probability of a non-random event, are common among many roulette tables. Additionally, a set of particular 'fail points' may result in a uniform response. Knowing what these individual fail points are, and how much each effect is to yield, is not necessary. It's the sum of failures that are important.

      Instead, one can perform principle components analysis from many, many datasets of various tables. From this, interactions between rolls and positions on the table can be estimated as vectors.

      When setting upon a target table, the several iterations of positions and roles can be recorded and converted into principle components, and a comparison of these vectors to the database may identify a comparable set of interactions. I think the remark about only requiring three rolls indicates the degree of correlation between sets of vectors gathered from tables with similar fail points.

      I used a similar scheme when I invented Polyatomic interference reduction spectroscopy (PAIRS), a method of determining interferences in ICP-MS. Essentially, interactions between various mass to charge ratios are elucidated as the relative concentration of interferences (polyatomic species, generally) is reduced relative to the elemental species. In this case, the identity of the individual interferences can be elucidated, via knowledge of chemical structure, addition of said interfences into a sample of known response, and similar. But this level of determination is not strictly necessary, if one is interested only in predicting future behavior based upon a run of limited data.

      I've applied this scheme to other instrumentation types and I believe it is applicable to many more fields than chemistry. If someone wants more information, send me a message.
    • I don't know the answer either, but it looks like you can calculate a lot [wikipedia.org].
  • Do the algorithms use the amount being bet on any particular choice to make that choice less likely to appear?

    Or are they programmed to favor the house only in that the payouts do not reflect the odds of a truly random outcome?

  • Today, most wheels have gone digital, run by algorithms programmed to favor the house.

    This makes me question whether the author has any actual idea what the F they're talking about.

    "Most" wheels certainly have not "gone digital" and are not "run by algorithms." Any game of purported "roulette" that has an algorithm other than randomness among available numbers picking the winner is not roulette. In the US, this is restricted to "Class II" gaming locations, where there's no randomness at all, but the player is being entertained by a digital lottery among all games connected together. Any Vega

    • by Anonymous Coward

      Or perhaps you're just reading way too much into the author's use of the words "algorithm" and "gone digital".

      Is software involved at any stage? Say, in regulating the amount of spin of the mechanical wheel, or controlling how the ball is launched? Then it's "gone digital" and involves "algorithms". Nowhere does the author suggest it's not still random - if anything, TFA says that old sources of predictability have now been eliminated.

      • Software is not involved in any stage other than an optical sensor (laser pickup) which picks up when a ball has settled into a slot.

        Roulette wheels have velocity added by hand and the attendant launches the ball themself.

        For games which *are* automated (and most aren't), the wheel is visivle and spins at a constant rate. Any attempt to time the launch of the ball by software would be HIGHLY illegal for any Vegas-style casino.

  • by AlanObject ( 3603453 ) on Sunday May 26, 2019 @06:20PM (#58659138)

    Today, if the Doc showed up and started consistently winning at the table the casino operator would quickly figure out why and have him shown the door. Politely and firmly. In most venues figuring out a way to beat the casino is against the law.

    The only casinos that are high probablility to lose money and go out of business are Trump casinos.

    • by hawk ( 1151 )

      > In most venues figuring out a way to beat the casino is against the law.

      It's certainly not against Nevada law.

      Similarly, the law doesn't require the casino to take the bets any more than you are required to place the bet, and it may indeed disinvite you from the premises.

      (using a counting or other *device* is indeed against Nevada law).

      hawk

  • At the time...computers were looked upon as creatures from outer space... Few persons, including casino managers, were vocationally qualified to distinguish myth from reality." Hiding behind this technological ruse...

    We still have such bullshit today: "I used blockchain quantum AI with web-scale microservices." Or just, "Watson did it."

  • by quax ( 19371 ) on Sunday May 26, 2019 @09:28PM (#58659950)

    Researchers from the Heidelberg compute center did something very similar using their IBM mainframe in the sixties to bet and increase research money at the casino in Baden-Baden.

    The story went that they eventually got busted and sent to jail although they didn't take any money for personal gains.

    I heard this story while being a graduate at the University of Heidelberg, but strangely I can find no traces of it online. So I cannot vouch for its validity.

    On the other hand, this was independently replicated in the seventies in the US: https://en.wikipedia.org/wiki/... [wikipedia.org]

  • Shorty dreams [americanliterature.com]
    -- Jack London, 1911

  • ... is not a word. Author of the original article is a moron for writing it, and Editor David is a moron for repeating it. Yet another sign that the "tech press" is mostly comprised of clueless mopes who held unspecified "IT jobs" before deciding that they wanted to write ignorant bullshit on the internet for a living.

When your work speaks for itself, don't interrupt. -- Henry J. Kaiser

Working...