Quote:
Originally Posted by AMEC0404
OK, Im good at math and ****, but let me get this straight... if we take my next 10K AIPF hands and I run exactly the same as my 10K AIPF hands already analized, that means 1 out of 324 people run that bad with that sample? Is that correct?
And then after another 10K, if I ran the same, 1 out of 5832 would run that bad over that sample?
Is that right?
What if the 2nd 10K hand sample, only 1 out of 9 people run that bad. Then when looking at both samples it would be 1 out of 162?
How high would the denominator have to be before a red flag goes up?
Well, I'll defer to the better stats people on that. I came up with those numbers, but your situation has to be discounted slightly because first of all, we're dealing with what is in essence a cherrypicked sample. The odds that your next sample of the same number of all ins is as bad as your first is simply 1 in 18. That sample has no memory of the previous sample.
However, if you said right now I'm going to play 10k more all in hands, end sample, then another 10k all in hands, end sample, then another and we know exactly what we're looking for going in. Then yes over your next 3 samples of that many hands It should be 1 in 5832.
The reason it has to be done this way, is because any of us can go back and find something in our games that has happened which we'd consider rare. Say I flopped a royal flush yesterday.... well, that doesn't change my odds of flopping another one. I can't just pick yesterday as my starting point for the sample in hopes that if my next one comes sooner than later, I've somehow found a crazy statistical anomaly.
And to answer your question.... this is why we need to continuously test. I mean there are thousands of tournament players out there. If something happens to you that is one in 400 and we're testing for it going in, I guess a red flag kind of goes up, but we'd clearly need more testing to verify it. With thousands of players out there, it has to happen to somebody and that somebody could be you. But if 1 in 400 becomes 1 in 6000, then 1 in 100,000, of course a red flag is going to go up and we're going to need to really dig into this stuff.
It probably still makes more sense to just look at the samples as a whole though. For example, I'm pretty sure that if we just looked at 1 million hands together we'd get more statistical certainty than we would arbitrarily choosing our sample sizes. There should be no difference between hand #10,000 and hand #10,001.