Quote:
Originally Posted by cjhmdm
Precisely why I stated the only way to prove one way or another would be for the poker rooms to open up and make hand histories public. And, we already know this will never happen, which also means we'll never be able to provide the proof the defenders of this argument are 'requiring'. So, all of us 'rigtards' are simply wasting our time questioning anything, and the 'pros' know this, and know that we are wasting our time; which is why they do nothing but 'defend' with 'math' they cannot explain, and call us rigtards.
And it wouldn't matter how many hands histories I could personally show... Even if I gave 100 trillion hands, the data would be incomplete and therefore broken down and explained away, with the hopes of convincing me that everything is on the up and up even though we both know I just gave you 100 trillion lines of incomplete data.
It does matter how many HHs you show. The thing is, that's not the only thing that matters. The other thing that matters is what exactly you're testing. These two things have to be looked at together. One without the other is meaningless. Consider this:
Suppose I said I played a little bit on some site last night and I'm convinced it was rigged or buggy or something. Someone might ask how many hands I played. If I said I played about 15 or 20 hands, a lot of people would dismiss it. But if I added that I was dealt the exact same hand, 3c2h, every time, the same people who dismissed it would now agree that something was wrong.
But if I said I played those 15 or 20 hands and didn't once get dealt pocket Aces and that's why I think it's rigged, then
everybody would dismiss my argument.
It should be clear that we can't just look at sample size to dismiss someone's argument. It should be equally clear that we can't just look at the probability of an event to make any kind of judgment about someone's argument. What we need is some number that considers both sample size and the probability of single instance of an event. If we had that, we should be able to use it to judge whether someone's argument really is valid and worth investigating.
As it turns out, there is such a number. It's called standard deviation. It's what Josem and others used to expose the cheating that went on at Absolute Poker a few years ago. They didn't have that many hand histories to work with, but they didn't need that many because what they discovered was something like being dealt the same hand many times in a row. I think Josem said that Potripper's performance was 15 standard deviations above the mean. That's about as close as you can get to ironclad proof that something was wrong.
The truth is, you don't need to find something
that outrageous. Spadebidder has said that he would consider anything 4 standard deviations from the mean to be worthy of further investigation, and that sounds like a pretty fair cutoff. There is a better than 99.99% probability that whatever you observe will fall within 4 standard deviations, assuming a fair deal. If it falls outside that, it's worth further testing to see if the deal really is fair. If a site is rigged, further testing will not merely increase the sample size; it will also increase the number of standard deviations the sample departs from the mean, and this will get you closer to proving that it's rigged. If the site continues to rig its deal in the way you're testing for, your sample size will eventually be big enough to prove beyond any reasonable doubt that it is rigged.
Somewhere between 6 and 7 standard deviations should be enough to convince almost anybody. The odds against seeing such a thing by chance are somewhere between 500 million to one and 400 billion to one.
tl;dr -- Sample size alone is meaningless. So is the probability of the event. Sample size and probability must be considered together to make a meaningful argument. You can use the standard deviation statistic to do this.