ScottC September 23, 2015 9:19 AM

I remember watching this live. At the time it was pretty amazing, and the show was so obnoxious that it was pretty fun watching them get taken to the cleaners. I had no idea he had such a con-artist background or that it ended up that badly.

Dave September 23, 2015 10:04 AM


Reading the article, it doesn’t sound like random luck to me. The guy obviously put a lot of time and effort into beating the system. That took skill, not luck.

The man was obviously a very clever scoundrel. But a scoundrel nonetheless. I suppose it’s poetic justice that the money ultimately didn’t do him much good.

dp September 23, 2015 10:24 AM

Ada Lovelace ruined herself gambling on horse races in the hope of building the first computer because she assumed she could beat the odds.

Steve Friedl September 23, 2015 11:38 AM

I gotta hand it to CBS for not trying to welch out on the deal, agreeing that they were beaten fair and square. I’m sure these days there’s be some clause that would somehow allow them not to pay.

root@localhost September 23, 2015 12:03 PM

I don’t get some of these comments. So maybe he wasn’t the nicest guy in the world.

There’s a reason counting cards isn’t illegal. Or being a poker mechanic.
If you don’t want to play with me, that is fine.

If you want luck, flip a coin. Even then it isn’t 50/50.
Know the odds, place your bets, and press your luck.

Matt September 23, 2015 12:06 PM

This story expands on Schneier’s advice about building your own crypto code.

Building your own pseudo-random number generator is not advisable. Especially if it is used for gambling or cryptography.

Winter September 23, 2015 12:21 PM

“I don’t get some of these comments. So maybe he wasn’t the nicest guy in the world.”

That is in some way relevant. Know thy enemy.

Bryan September 23, 2015 12:33 PM

A relative of mine built that show’s board. I don’t know if he built the randomizer that Larson pwned, but I suspect he may have. Which is definitely something to keep in mind the next time I design anything that… well, anything at all.

Dave. September 23, 2015 1:03 PM

@Steve Friedl,

These days he’d get a visit from the FBI and get thrown in jail for “Computer Fraud”.

Dave September 23, 2015 2:27 PM

From what I recall, the sequence wasn’t random, it was a repeating pattern with several state changes per second. What the guy did was figure out just when to press the button to get the “Get some money and go again” square (instead of the much more common “Get some money” or “Lose your money” squares). He had to press randomly to end his run, as all he practiced for was the “Get some money and go again” result. Fortunately (for him), he got a “Get some money” square, not a “Lose your money” square.

Milo M. September 23, 2015 2:28 PM

The Priceonomics story reads a lot like the Wikipedia entry on Larson:

But then the Wikipedia entry is based in large part on a 2003 episode on the Game Show Network.

And the wheel goes round and round.

Looks like one could eke out a living rewording some of the longer Wikipedia articles.

rgaff September 23, 2015 4:02 PM

@ Dave

A computer that randomly picks heads, and then from that point onward replays the same “random” heads result over and over still picked heads randomly… just the implementation misused that random result and converted it into something quite predictable…

Likewise the sequence was random, just it kept repeating the same “random” sequence over and over… technically, it says there were 5 different random sequences, though I’m unclear whether those 5 sequences were in random order or if the order was repeating…

Clive Robinson September 23, 2015 4:04 PM

Hmm another mysterious “clive robinson”… not the first time one has popped up.

Obviously not me as I don’t make comments that short unless they are “one liners”… 😉

@ Dave,

Sorry not me.

But whilst a small amount of luck was initialy required whilst he “learned the button”, from then it was not just physical skill it was also his ability to remember the sequences to work out quickly which one he was playing with.

Whilst this is the first known “TV” breaking of a “gaming PRNG” I’m not sure it was the first actual break. Various Casino Games using machines have been hit over the years, some mechanical many electronic. In most cases the attacker has been able to obtain a machine and manual and work it out in the luxury of their own private space.

The big problem I’ve been told is the “House percentage” in that it’s supposed to pay at a constant rate compared to play, which also makes the Win payout rate much more constant, further the machine can not usually pay more than it holds in the win buffer and for some machines in non 24×7 venues, the amount that can be held in the machine is also limited by insurance rules. All of which degrades the design of the PRNG, which has led in the past to it being a little to predictable.

65535 September 23, 2015 8:22 PM

Someone check those Whammy numbers again. If he had 5/6th of a chance per spin of not hitting the Whammy is the calculation correct? And, given the fact he could hit square 4 and 8 correctly each spin [and thus avoid the Whammy for the most part] should not the calculation be more complex?

“Given the 1 in 6 odds of hitting a “Whammy,” the probability of going 45 spins in a row without hitting one was (5/6)^45, or .027%. Larson had beat odds of roughly 3 out of 10,000.” –Priceonomicis

MrTroy September 24, 2015 12:29 AM


The article is talking about the naive probability of making 45 spins without hitting a whammy if not using any tricks. The fact that the odds are so low help to suggest that a trick may have been involved.

Peter A. September 24, 2015 3:18 AM

From what the article says (and from a casual review of the show recordings on YouTube) it is doubtful that it was a – shockingly – bad PRNG fault. 1980’s PRNGs were quite OK, even my ZX Spectrum had a full-cycle 16-bit congruential PRNG built into its BASIC interpreter. Even a mediocre 8-bit full-cycle PRNG would be next to impossible to memorize its sequence. In fact, I doubt there was any PRNG involved in the original design of the show’s device, unless you call this a PRNG.

Article says explicitly there were five fixed sequences of the “jumping frame” positions, however it does not elaborate on the sequences’ length, it looks like it was about 10-20. Maybe one of the five sequences was chosen at random before each “spin”, but I doubt it. The article suggests that the original device was a fully hardware one (it says that “a PC” was added later). I would expect a short linear feedback register at best.

In my opinion, the fault lays square on the design: 1. trying to appeal to humans’ bad perception of randomness, 2. ignoring human’s brain ability to detect patterns and 3. underestimating human eye-hand coordination skills.

  1. The sequences, even if fixed, were not generated by a (P)RNG. They were doctored to look random for a human, by making that frame “jump” a long distance on subsequent positions and never “stay” on one square. True- or good pseudo-random sequences occasionally contain “streaks” of the same value (if the generator allows it, congruential ones do not) or values close to each other – which is falsely seen by many as “non-randomness”. For instance, if a particular “X out of Y” number lottery draw contains subsequent numbers, some people cry “fraud” on the house – but the probability of such a subset being drawn is the same as any other.
  2. Humans are quite good at detecting patterns and cycles, even “random” ones; and even non-existing ones 🙂 The designers of the game have ignored or underestimated this ability and also the human ability to rehearse and develop nearly automatic reaction to repeating complex patterns. Examples: martial arts, fencing, fast-paced games like table tennis etc. etc. Apparently it was thought that a fast changing pattern was next to impossible to determine. They’ve also ignored existing helpful consumer technology – a video recorder/player with still image function 🙂
  3. Human brain is hard-wired for visual motion detection and predictive expedited muscle control that is called eye-hand coordination. The neural signal propagation is quite slow, therefore the brain has to “send” the “control sequence” to the muscles quite ahead of time, yet it is able to move the limbs to meet a fast-moving object (a ball etc.) in the right nick of time to achieve the desired effect (catching, bouncing in the desired direction etc.) Apparently, the designer(s) thought that the interval in which the frame stayed on every single square is too short to react with button press – and it is! But Mr Larson did not do that – he recognized the pattern and knew that the frame is coming to the winning square ahead of time, and moved his hand ahead of time to press the button at the very nick of time in which the frame arrived where he wanted. It’s like learning to catch a ball. He practiced it at home and during the show had only to adjust to the button’s “feel”, i.e. the delay it introduced – mechanically or electrically.

The final fault of the show’s designers was creating “always winning” squares with only “get some money and get to try again” prizes on it.

All in all, it was a perfect and completely fair Larsony.

65535 September 24, 2015 6:40 AM

@ MrTroy

I see what you are saying. It’s a raw percentage assuming no knowledge.

But, it was game and was on the air for some time [long enough to video tape enough shows to see the five loops]. One of the other contenders must have done some studying also before entering [in addition to Mr. Larson].


“Real-life roulette exploits
• In 1873, Briton Joseph Jaggers made the first famous biased roulette wheel exploit. Mr. Jaggers, with a team of six accomplices, carefully observed all the wheels at the Monte Carlo casino and found one wheel with significant bias. By taking advantage of this flaw they managed to win over $325,000, an astronomical sum in 1873.
• In the summer of 1891 at the Monte Carlo casino, a part-time swindler and petty crook from London named Charles Wells broke the bank at each table he played over a period of several days. Breaking the bank meant he won all the available money in the table bank that day, and a black cloth would be placed over the table until the bank was replenished. In song and life, he was celebrated as “The Man Who Broke the Bank at Monte Carlo”.
• In 1992 in Gdynia (Poland) in Jackpol Casino, Paweł Piskorski (Polish politician, former Secretary General of Platforma Obywatelska and Member of European Parliament, today Leader of Stronnictwo Demokratyczne) claimed to have won 4,950,000,000 zloty (approximately $175,000 today). This, at least, was the explanation he gave to the revenue office for his income that year. The claim is implausible, because the house maximum at that time was 1,000,000 zloty: he would have had to have won 138 times in succession, with each bet set at the house maximum. The odds of hitting a winning number 138 times in a row is approximately 1 in 5.88×10^214.
• In 2004, Ashley Revell of London sold all of his possessions, clothing included, and placed his entire net worth of US$135,300 on red at the Plaza Hotel in Las Vegas. The ball landed on “Red 7” and Revell walked away with $270,600.
• On 2 October 2009, Derren Brown (as part of his controversial “The Events” series) bet £5000 of a member of the public’s money on a single number of a roulette wheel somewhere in Europe. This was shown live across the UK using a camera hidden in Brown’s sleeve. He claimed that he would use the laws of physics to predict where the ball would end up, based upon the speed of the wheel and the ball. Brown took approximately three seconds after the wheel started spinning to place his £5000 bet on the number 8, only to see it land on 30 — just one number out. It is not known whether any other less successful bets were filmed but not broadcast (Brown has used, and explained, this “cherry picking” technique on at least two other TV programmes, in coin tossing and horse racing).”


Several software packages exist for running computer-based simulation modeling (e.g. Monte Carlo simulation, stochastic modeling, multimethod modeling) that makes all the modeling almost effortless.

Surely, CBS must be familiar Monte Carlo situation in real life games. Nobody is going to blindly travel from the Midwest all the way to LA without some sense of gaming the game – so to speak.

Clive Robinson September 24, 2015 9:51 AM

@ Peter A.,

From what the article says it is doubtful that it was a – shockingly – bad PRNG fault.

What would you call it?

As far as I’m concerned it’s “a sequencer with a pseudo random output mapping”, which is the fundemental model for all PRNGs even CS-PRNGs. The fact that it always starts in the same place and has an apparently short run time does not make it any less a PRNG.

You say,

Even a mediocre 8-bit full-cycle PRNG would be next to impossible to memorize its sequence.

If you mean “it’a FULL sequence” maybe, people have and do memorise the order of more than four packs of cards, given a realitivly short period of time. But if an 8bit PRNG is used, if it always starts in the same place, then to work this trick he would not have had to memorize anywhere near the full cycle. Maybe just one or two “lead in” positions to identify the sequence being used, then maybe three to eight positions –similar to the way a matched filter works– to trigger his button push point.

r September 24, 2015 11:24 AM

@steve friedl & dave

i believe in vegas these days if they determine you to be manually counting cards they will throw you out, i’m not sure as to whether it is truly illegal but… this is what i believe is the modern day equivalent.

i say he beat the house at their own game. 🙂
just like what mr. schneier says about non-renouned ‘cryptographers’ releasing cryptographic code… it’s just not a very good idea.

Fred P September 24, 2015 1:44 PM

This reminds me of an old timing exploit on mechanical slot machines – apparently someone who pulled the handle at exactly the right time could guarantee the value on the first wheel or (with a lot more practice) two. (Source: Scarne’s New Complete Guide to Gambling) This is why in a modern slot machine, the timing that you hit the button/screen/pull the handle has nothing to do with your result.


  • When I first read about it, I had great difficulty understanding what this “counting cards” trick was, and how I could do it. After a few years, I found out that it was merely using the available data to understand the current odds better, which I’d been doing since I was 7, if my ability to use it effectively was low at first.

I mean, really, what’s the point in playing a card game if you aren’t counting cards?

Anonymous Cow September 24, 2015 2:28 PM

…in vegas these days if they determine you to be manually counting cards they will throw you out, i’m not sure as to whether it is truly illegal…

Correct: card counting is NOT illegal. BUT: the casino can exercise it’s rights as a business owner to refuse service to anybody, and ask anybody to leave the property. I don’t know if Nevada has the same laws as my area, but refusing to leave a property when requested to do so can be prosecuted as burglary.

tyr September 24, 2015 10:44 PM

@ Fred P.

The mechanical slot machines had a friction clutch
assembly to vary the amount of handle torque needed
to duplicate the handle pull. The way to learn how
was to have one with exposed interior to practice
on for awhile. Turned out to be a useless skill in
the age of electronic slot machines.

JD September 24, 2015 11:05 PM

Mechanical games might have had tricks, but in these days of video slot games, any “advantage” that you could think to hack usually doesn’t even exist. A programmer who’s worked on these, Voja Antonic, tells of his experience below. On one poker game, at the moment that you press the button to show your card, the card isn’t actually computed until after you press the button! “I witnessed a lot of players who, when they would lose, shouted ‘Damn, why didn’t I press…’. How could I tell them that the outcome would have been the same?” (Though, with his experience, he does note some current vulnerabilities such as with optical readers.)

Jon September 25, 2015 2:46 AM

Just an aside: The wheels in a modern one-armed bandit are driven by stepper motors, and yes, the internal processor, if it determines the machine has paid out too much lately, can and will add another step to the spin, changing a win to a loss.


Zero October 21, 2015 6:08 PM

Re: Counting cards.

You can be trespassed in Nevada. A while back I read that most casinos in Nevada let you win about $0.25 per hand without harassing you. In Atlantic city they’re legally not allowed to kick you out for counting so they play with different odds that make it not worth it financially to play blackjack.

Leave a comment


Allowed HTML <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre> Markdown Extra syntax via

Sidebar photo of Bruce Schneier by Joe MacInnis.