Quote:
Originally Posted by Lestat
Correct.
I don't understand... Do you mean show technically how to tamper with the RNG? That's way beyond my pay grade. Otherwise, I'm not sure what you're saying. I thought I presented a theory and showed why if true, it would be almost impossible to detect. I'm really not good with this stuff and others here are a lot smarter than me. I'm just saying that if you tweaked equity percentages ever so slightly, and further fine tuned it so that it's more prevelant in the rarer large pots, than in the much more common smaller ones, combined with not being privvy to folded hole cards, I don't see how anyone could detect that. If you're asking me to provide details on how all this is implemented from an engineering standpoint, then I'll just concede now that I'm full of hot air. I'd have no idea where to start.
The problem with that idea is that no matter how small you make your tweaks, it will still be detectable as the number of hands in your sample grows. Let's do a concrete example:
Suppose we want to tweak dominated hands slightly so that the better hand didn't win quite as much as it ought to. AK vs A[Q or lower] averages out to around 73-27 in favor of AK. We don't want to be too obvious, so let's just make it 70-30. The AK still wins the large majority of those matchups, but the AQ wins a few more than he ought to, hopefully leading to him occasionally staying in the game longer and contributing more rake. (I'm assuming that's the point of this tweaking.)
After 100 hands, the AK has won 70 times. He expected to win 73, but 70 is pretty close, and you wouldn't expect this to cause any alarm, right? Actually, though, there was only a 28% chance that he would win 70 (or fewer) times. Still not an outrageous result, though. We see 28 percenters hit all the time. No big deal. Of course, only three more hands than expected were won by the underdog. That's not a very big addition to our rake.
But one of the reasons tweaking such a small percentage seemed like a good plan is because there are so many hands played online, we know it will add up pretty quickly for us. Unfortunately for our plan, so will the error.
After 1000 hands, when the AK has won 700 (instead of 730), things start getting a little less plausible. There is a 98.1% chance that AK will win more than 700 times out of 1000 against a dominated hand in a fair game.
As our added rake piles up, so does the error we've created. After 10,000 hands in our rigged game, AK has only won 7,000 times. The probability that AK will win more than 7,000 times in a fair game against a dominated hand is:
99.9999999987532%
I think you'll agree that this would be flagged by anybody running any kind of analysis on the hand histories. And we've only flipped the results of 300 hands. Not a huge addition to our bottom line, no matter what the winners decide to do with their ill-gotten pots. It appears that the error grows much faster than any difference we can make to our profits, and anyone looking for anomalies in the data will be able to spot them by the time we've made enough to pay for dinner and a movie.
It should be clear from this example that no matter how tiny the tweak, the error will grow so quickly that it will overwhelm our additional profit. In real life, it won't be noticed quite this quickly because some players will fold AQ in some situations and because no one has access to complete information about every hand played on any site. But even with incomplete information and not every hand being played to completion, the pattern will still reveal itself through careful analysis.
Periodically changing which situations you tweak will accomplish nothing, since with this type of analysis it doesn't matter if it's a dominated hand winning too much or a gutshot hitting too often or KK beating AA more than it should. The only thing that matters is the discrepancy between expectation and result.