NoahSD's Awesome Poker Blog

Eight days ago, I wrote a post about Absolute Poker’s ridiculously non-random Keno, which detailed a pathetically incompetent mistake that they had made (or perhaps that an outside contractor, Betsoft Gaming, had made that they’d completely failed to notice). It also explained that their official explanation was a lie and that over five months had gone by without compensation or a better explanation. (I highly suggest reading that post before this one. Otherwise, you’ll have absolutely no clue what I’m talking about. Plus, it’s worth the read.)

Well, I’ve been paid back. At 2:00 today, I got this e-mail from AP (I bolded the important part):

Dear Noah,

As a valued customer and Keno player at our site, we want to update you on an issue we were made aware of that suggested the Keno game we offered to players was operating in a non-random way.

Our early analysis of this claim stated that there was a flaw specific to the game logic in ‘ten play’ only on Traditional Keno and no confirmed issue with the Random Number Generator (RNG). Upon notice of the flaw, we contacted our third party software provider of Keno, Betsoft Gaming (BSG) for a more comprehensive investigation and subsequently removed all Keno games from our site until the investigation was complete and all corrective actions taken.

Our joint investigations with BSG confirmed that while the RNG for Keno was functioning properly, it was selecting from a pre-existing data set that was not as extensive as it should have been. The results were in turn random, but from a limited data set rather than from the full data set from which RNG is expected to choose.

While the issue stemmed from our third party software provider, we accept responsibility for offering our players a Keno game that had the described issue. Therefore refunds have been issued to all Keno players of ‘five Play’, ’ten Play’ and all Traditional Keno during the period those games were offered . Any net loss you experienced while playing Keno with us, no matter how small has been credited to your account. [emphasis mine]

Your account has been credited with:

Total Amount: $12.70

We wish to offer our sincere apologies for any confusion or frustration this issue may have caused you, and we thank you for your patience while we have endeavored to correct this situation. We also want to restate our commitment to providing a safe, secure and fair gaming environment for our players.

If you have any questions about your refund, please write to us at [removed so they don’t get spammed]

Thank you for your understanding, support and continued confidence in our game offerings.

The Absolute Poker Team

I’ve checked with a few people, and everyone seems to have been repaid a reasonable amount. (Obviously, if you haven’t been repaid, please get in contact with me.) I got my $12.70 from my degen Keno investigation, so that’s cool.

They calculated payments fairly enough. Even though their initial statement said that the problem was only with the ’10 play’ feature (which was a lie), they say they’ve repaid for all play. They say they used net losses from all those games, which I think is perfectly reasonable. I didn’t keep track of how much I lost while investigating this, but the $12.70 seems about right.

So, they deserve credit for eventually repaying their customers.

I’m quite tempted to end my post there because I hate to attack them in a post that I started writing because they repaid their customers. They should be thanked for repaying their customers, and they should certainly not be discouraged from doing the right thing in the future by a fear that any action that they take will lead to negative press. I swear that had they handled any part of this decently, I would have been much much more forgiving given that they did repay customers in the end. But, they screwed up so much here and revealed so much incompetence, negligence, and maliciousness that I just can’t in good conscience bury the hatchet.

First, Remember How We Got Here

The only reason that they needed to repay their customers is because they had this rigged game up on their site. That’s obviously something that should never happen.

And of course, they took forever to repay. Five-and-a-half months after myself and Thomas Bakker made this public, they’ve just now gotten around to repaying. That also should never happen. And, while I obviously can’t prove that they only decided to repay because I made a stink about it and got them some bad publicity, the fact that they didn’t repay for so long definitely makes it look like they weren’t planning on repaying. (Their statement, as relayed to me by Joe Sebok, that they had been waiting for BSG to provide data is obviously ridiculous. The money came from AP player accounts to AP’s company coffers. They of course have an internal transaction history of all of that.)

(I don’t know how much they ended up repaying, but I think it’s pretty safe to assume that it’s a relatively insignificant amount for such a large company. For example, Joe Sebok said at one point that he “believed” it was under $30k. I’m not sure if that makes it better or worse that they’ve taken so long to repay it.)

Their Explanation Reveals Incredible Incompetence and Sketchiness

Much more importantly, their explanation is still a lie, and it shows an amazing level of incompetence. See this part of the e-mail:

Our joint investigations with BSG confirmed that while the RNG for Keno was functioning properly, it was selecting from a pre-existing data set that was not as extensive as it should have been. The results were in turn random, but from a limited data set rather than from the full data set from which RNG is expected to choose.

Translation: They’re saying that they had generated some set of random spins of Keno (or maybe random sets of ten spins for their ’10 play’ feature) and then just randomly selected from them.

First of all, that is not how a casino game is supposed to run.  You don’t randomly select from some results that you generated earlier. You randomly select from all possibilities with the proper weightings. That should be completely obvious. And, doing it the way that they say has a lot of flaws. In Keno, it would have the one major flaw that a customer’s win distribution isn’t the win distribution expected from the game; it’s whatever win distribution the game designers got from their inital generated set. That’s wrong already because customers are signing up to play a random game of Keno and not a game of “let’s pick randomly from the results that the game designers got.”

So, yeah, that’s bad. But, use of this method gets pretty sketchy when you consider that sets can be regenerated: Suppose that the game designers generated a set that happened to result in worse-than-expected results for the house. That obviously happens about half the time, and if AP notices this and AP decides to “correct” this when it happens, that would constitute rigging of the game in the houses favor–The game would always be slanted, but never significantly in the player’s favor. The really terrifying thing about this is that the game could be rigged in the house’s favor as a result of nothing more than an innocent and incompetent employee thinking “Hmmm… This dataset isn’t good. I’m going to generate another one.”

Plus, this would be essentially undetectable. Even if we could somehow get a massive sample of data (or even the originally generated set itself), we wouldn’t be able to distinguish between a randomly generated dataset that happened to give the house an advantage and one that was chosen (either due to maliciousness or incompetence) specifically because it gave a house advantage.

How big of a deal this is obviously depends on how extensive the dataset was, since larger datasets are much less likely to be slanted much in any direction. This quote from the e-mail might be relevant: “a pre-existing data set that was not as extensive as it should have been.” So, while I don’t know specifically that this happened in this case (Over the small sample where I played this game and actually recorded my results, I actually ran better than expected), this method of generating data requires that AP’s employees be both extremely competent and of very high integrity or else their games could be easily rigged in the house’s favor. That would be a ridiculous amount of trust to put in any company. It’s absurd for AP to expect that of us (though, to be fair, they may not have realized how flawed this was).

In other words, their current official explanation is “We were trying to set up a game in a way that wouldn’t provide our customers with the game that they wanted and could allow for easy (and even accidental) manipulation by our employees, but we didn’t do it right.”

This actually isn’t the only example of AP using this methodology.

In fact (and I can’t help but bring this up, because Jesus I can’t believe AP is actually admitting to this publicly), this is not the only instance of Cereus using this incredibly stupid and sketchy methodology. (Sorry for the vagueness. I’m sitting on a fairly big story, but I just can’t tell it until I’ve fleshed out exactly what happened. Suffice it to say that the people who work for Cereus who are reading this know what I’m talking about, and you will soon.)

(I should say that I’m using AP here as a catch-all for AP, BSG, and TST.  It may well be the case that Betsoft Gaming made this ridiculous game and not AP. So some of the things that I directly attribute to Absolute Poker could have been done by BSG instead. However, AP put this on their web site so they’re responsible for it. They’re a massive online gambling company, so they should be expected to know how these things should be run and to check that they are run correctly before putting them on their web site. I sympathize in general with the unfortunate consequences of hiring incompetent outside contractors, but AP could have done any tiny amount of due diligence and discovered that something was up. Hell, this was discovered by someone who was just playing the game casually and had no access to anything behind the scenes.

And, of couse, the company Technical Systems Testing that certified this game as random is obviously just a rubber stamp. There’s no way that anyone could be hired for the explicit job of checking the randomness of this game and conclude that it was okay unless they simply didn’t do what they say they did.)

Also, It’s a Lie

So, yeah, that explanation–the one that goes “We were trying to do things in the dumbest and sketchiest way possible, but we couldn’t even get that right”–That’s not true. How do I know? Because it’s completely inconsistent with what happened.

The flaw that I was shown and that Thomas and I presented to you in our original video showed a very clear pattern: When you picked five numbers and ran ten consecutive spins with the ’10 play’ feature, there was an absurdly high likelihood (86%) that the nth and (n+5)th spin would hit 3 numbers and no other round would pay. (That makes a lot more sense when you actually watch the video, which is the first one in my previous post on the topic.)

So AP is now claiming that they randomly generated a dataset that had those absurdly predictable results? That doesn’t explain anything. All it does is leads to the question “Why did your ‘randomly’ generated dataset have such an incredibly unlikely pattern in it?” I already did the math in my original 2p2 post from 5 months ago that showed that over a sample size of only 50, the odds of seeing such an absurd pattern are about one in 10^45. So, unless their “pre-existing data set that was not as extensive as it should have been” actually had less than 50 data points in it, this just isn’t true.

(I acknowledge that this quasi-mathematical argument is slightly hand-wavy and disingenuous, but I’m hoping the fellow nerds in the audience will give me a pass here. The point is clear, and the actual math is tedious. I just don’t have the time to dot every i right now.)

In fact, there’s tons of other evidence that their new explanation is a lie (both things that I made public five months ago and things that I’m keeping to myself for now). But, AP has shown that they’re willing to lie about this, and they’ve also shown that they’re very bad at it. Therefore, I will not make it easier to lie by spelling out exactly which explanations make sense and which don’t. They can feel free to try again, or they can just actually say what happened.

More likely, they’ll just drop it now that they’ve repaid.