[ QUOTE ]
Using Google, I tried to find something else that occurred at fifteen standard deviations. I couldn't find anything.
The
Wikipedia entry on Standard Deviation lists the confidence interval of 7SD at 99.99999999974%. That is to say, that on a normally distributed data set, 99.999999999,74% of the results will be within standard deviations - or, to put it another way, 26/100,000,000,000 - twenty six in every hundred billion.
[/ QUOTE ]
While I think the evidence is exremely strong, this greatly overstates the case, since nothing says that the distribution has to be normal or anywhere close to normal. The Central Limit Theorem says nothing about the underlying distribution of true win rates. You can't use the probability that a normally distributed random variable would be 15 or more standard deviations away from average.
More work needs to be done to quantify how unlikely it would be to see something like this when you can't see your opponents' hole cards. One problem is that there could be other tells that would allow a non-cheating player to get information. For example, if a player thinks for 20 seconds and then checks, this might mean they have a different hand range than if they check quickly, and someone who just recognizes this difference may appear to have inside information according to a naive test.