Do these results warrant an investigation? -4stdev in 646 bets

I played 3 sessions 3 different websites with the same software provider. My results are as follows:





How suspicious are these results? Who could I contact to investigate? I assure you these results are genuine and have video + hand histories to prove it.

(mod note: edited to display graphics, rather than links in thread)

Suspicious, but on their own not enough evidence to do anything about. The next step is hypothesize how the game is treating you unfairly, and run a statistical test on the associated card values. You can check your total, the dealer's up card, down card, 3rd drawn card, etc. These are all known ways to tamper with a blackjack game.

I think 646 bets is not enough to get a true reflection of the game. 4 std deviation confidence limit can happen with such a small sample size. You need thousands of hands in order to be sure the game is not treating you fairly.

What does "Chance of Loss" mean ? Then chance that you occur precisely the loss amount you entered into the field ? This number will always be ridiculously low.

Or is this meant as "entered loss or worse" ? Then I don't see how this will ever be significant. You made the test *because* you lost heavily after a few hands. Hence all possible analysis results on those sessions will be heavily negative, and the chance of occuring will be way higher than indicated.

Quote: MangoJ

What does "Chance of Loss" mean ? Then chance that you occur precisely the loss amount you entered into the field ? This number will always be ridiculously low.

Or is this meant as "entered loss or worse" ? Then I don't see how this will ever be significant. You made the test *because* you lost heavily after a few hands. Hence all possible analysis results on those sessions will be heavily negative, and the chance of occuring will be way higher than indicated.

The data was entered here:

The chance of loss means (I assume) the probability to lose X amount or worse.

Quote: roznez

The data was entered here: http://www.beatingbonuses.com/calc_fair.htm

The chance of loss means (I assume) the probability to lose X amount or worse.

It looks like there is nowhere there to enter the blackjack rules, so I have no idea how they come up with a probability of losing.

Also, it probably assumes that you play perfect, mistake-free blackjack, which (for most people) is not true.

Even if the numbers are accurate, they don't really mean much on their own. Unlikely events happen. This is probably enough hands that your possible results more or less look like a normal distribution, so the 4SD number is meaningful.

The problem here is the selection bias. If the probability of losing this badly is 1 in 20,000, that means that it is supposed to happen sometimes -- it's supposed to happen to 1 in 20,000 people. Of course, those are the people who complain (the guy who is 4 SD above the mean is not posting to a message board asking if the game is fair)

Which is not to say that the game is fair -- just to say that I don't know. Certainly, many online casinos cheat; this could be one of them. Have you searched around for the casino and the software to see if there are an unusually high number of complaints?

With respect to the BeatingBonuses simulator, if it is like the others I've seen, you don't input the Rules, you simply input the probability of each possible result and what it pays. Obviously, the Rules themselves will change the probability distribution of results, so one must make sure one has the probabilities corresponding correctly to the set of Rules.

Assuming the OP has done this, and I think this has been mentioned before, but how he lost might be even more important than that he lost. In other words, the probability of him losing this badly can be 1 in 20,000, but he could have a bust rate that's even worse than that, or there could be the dealer bad beating him to such extent that the probability is much lower than that.

I would say the easiest ways to rig a Blackjack game (and I know nothing about programming, but I do know about 'friendly' house games) would be dealing from the bottom (known card) to either the player or the dealer, whichever is chosen. The equivalent to this on the Dealer side, for example, is to toss out the first bust card and go with the second card, when drawing, even if the second card would bust you. Same thing with the Player, when drawing, toss out the first card that does not bust the player.

Quote: roznez

I played 3 sessions 3 different websites with the same software provider. My results are as follows:





How suspicious are these results? Who could I contact to investigate? I assure you these results are genuine and have video + hand histories to prove it.

(mod note: edited to display graphics, rather than links in thread)

I can't draw any scientific conclusions, and as other posters point out, there is some obvious missing data and some evident bias. However, I would feel it is a very safe guess that the casino is cheating. Are you going to continue playing until you can prove it with more solid evidence? If I ran an online casino, you wouldn't need a sim. You would just need my confession: I would cheat. In other words, it's better to presume they cheat absent contrary evidence.

Quote: Mission146

I would say the easiest ways to rig a Blackjack game (and I know nothing about programming, but I do know about 'friendly' house games) would be dealing from the bottom (known card) to either the player or the dealer, whichever is chosen. The equivalent to this on the Dealer side, for example, is to toss out the first bust card and go with the second card, when drawing, even if the second card would bust you. Same thing with the Player, when drawing, toss out the first card that does not bust the player.

Once you come up with such a scenario, it is quite simple to test for it and rather easy to detect. Of course you would need new data for this.

Although you could in principle cheat without detection (on singular events), you cannot do it consistently.

Quote: MangoJ

Although you could in principle cheat without detection (on singular events), you cannot do it consistently.

In my ethics free world, when I decide to cheat any player entering my online blackjack haven, all I do is program the player to get 19 on every 100th hand while the house gets a 20 on that hand. I am guessing there is not a single player who would ever be in the least bit suspicious, and I just added significantly to the house edge that you would have calculated given the stated rules of the game. Bleed the cow, don't kill it.

As far as the OP, the great likelihood is that it is a rigged game, and my guess is if you want to throw another $10,000 into their coffers, you could prove it.

Quote: AxiomOfChoice

It looks like there is nowhere there to enter the blackjack rules, so I have no idea how they come up with a probability of losing.

On the left hand side, you can select the software and the specific game. I blacked them out in my screenshot so as not to expose the site/game, yet.

screenshot of how the inputs vary between software & game:

Quote: AxiomOfChoice

Which is not to say that the game is fair -- just to say that I don't know. Certainly, many online casinos cheat; this could be one of them. Have you searched around for the casino and the software to see if there are an unusually high number of complaints?

This occurred on a very reputable site, but they have had security issues in the past.

Quote: Sonuvabish

I can't draw any scientific conclusions, and as other posters point out, there is some obvious missing data and some evident bias. However, I would feel it is a very safe guess that the casino is cheating. Are you going to continue playing until you can prove it with more solid evidence? If I ran an online casino, you wouldn't need a sim. You would just need my confession: I would cheat. In other words, it's better to presume they cheat absent contrary evidence.

Can you elaborate on missing data and bias?

There's no selection bias-- these are the first 3 sessions I played on this particular software client.

All the hand histories logged if anyone would be interested in taking a look.

The potential selection bias is inherent in the fact that you are the one posting. If 10000 people play this game over a few thousand hands, there is a good chance one person is going to do terribly. Then come here and complain about doing terribly. This is self-selection bias.

I'm not saying that's what happened here, as I mentioned above I find the numbers suspicious. This site has busted cheating casinos, and the threads start out exactly like this. But it's also possible you just got really unlucky.

As for the missing data, it is easier to make the case that a casino is cheating if we have complete hand histories. Then we can run statistical tests on cards in certain positions. With enough data, we can prove a casino is cheating (or not) with our mathemagic.

Yes, I would be interested in taking a look at the hand histories, thank you!

I'm mainly interested in hands where you took a hit with a bustable total and hands where the dealer took a hit with a bustable total.

Quote: roznez

Can you elaborate on missing data and bias?

There's no selection bias-- these are the first 3 sessions I played on this particular software client.

All the hand histories logged if anyone would be interested in taking a look.

There are no rules input. And it assumes a strategy that may have various departures from your own. I do not think these issues invalidate your data, they just make it imprecise. Maybe you have accounted for them in some way, I don't know. I am not suggesting you attempt to collect more data.

Quote: Mission146

Yes, I would be interested in taking a look at the hand histories, thank you!

I'm mainly interested in hands where you took a hit with a bustable total and hands where the dealer took a hit with a bustable total.

Can you send me a PM please? I'm not sure how to do it on this forum.

Quote: Sonuvabish

There are no rules input. And it assumes a strategy that may have various departures from your own. I do not think these issues invalidate your data, they just make it imprecise. Maybe you have accounted for them in some way, I don't know. I am not suggesting you attempt to collect more data.

My hunch is the beatingbonuses calculator computes numbers assuming the same strategy it recommends on the website. The strategy is = to the one posted on wizardofodds.

PM Sent.

I would compare the 3rd card when needed for DEALER and PLAYER. DO NOT COUNT Aces as the third card. The rest of the shoe has an average value of 7. Aces are different, as they score 1 or 11 as needed. However, the occurrance of an Ace should be 1 in 13.

Quote: roznez

My hunch is the beatingbonuses calculator computes numbers assuming the same strategy it recommends on the website. The strategy is = to the one posted on wizardofodds.

You are operating on a hunch, while assuming you made no errors. I would call this bias. The possible bias in the data could go in the opposite direction, making the loss likelihood appear higher than it actually is (and it is already extremely low). I was mainly addressing other comments. I don't think this is significant to the conclusion.

I'd like to get a bet down on Badung being the next nuked user at whatever the current odds are