Re: Probabilty of PT notes
Depends on what the statistic is. Take VPIP for example. If your opponent's true VPIP is p, then after n hands his sample VPIP (the thing you have in PT) has a binomial distribution with standard deviation sqrt(p*(1-p)/n).
Let's try plugging in some numbers. Say his true VPIP is .2, and you have 100 hands of data on him. Then the SD of your sample statistic is sqrt(.2*.8/100) = .04. This means that 95% of the time you will observe a VPIP between .12 and .28 for a player whose true VPIP is .2, given 100 hands of data.
After 300 hands the SD drops to about .02, so you'd expect to observe between .16 and .24.
Practically speaking, you can conclude that a few hundred hands is a reasonable sample size for getting pretty accurate estimates, and even 100 hands will give you a ballpark figure.
gm
|