How accurate is my BB/100 after x hands?

[Ludanto]

All the time one can read on the forums about people posting their game results and other people saying things like "that's not big enough of a sample to say anything".

Let's assume that my style of play doesn't get better or worse. Isn't the accuracy of the BB/100 from Poker Tracker related to the amount of hands I have played?

If the answer is yes then what is the formula to find out how close my PT's BB/100 is to my REAL BB/100? Is there a formula which I can use and then say "Because I have played x hands I can say that my REAL BB/100 is not farther away from what PT says by x BB/100!"?

[DWarrior]

Search for "confidence interval"

also check out http://www.castrovalva.com/~la/win.htm next to the black squares, the WR is the confidence interval for your winrate.

If you don't get it, read up on confidence intervals on wikipedia

[BruceZ]

[ QUOTE ]
All the time one can read on the forums about people posting their game results and other people saying things like "that's not big enough of a sample to say anything".

Let's assume that my style of play doesn't get better or worse. Isn't the accuracy of the BB/100 from Poker Tracker related to the amount of hands I have played?

If the answer is yes then what is the formula to find out how close my PT's BB/100 is to my REAL BB/100? Is there a formula which I can use and then say "Because I have played x hands I can say that my REAL BB/100 is not farther away from what PT says by x BB/100!"?

[/ QUOTE ]

First compute the standard error of your win rate SE = SD/sqrt(N/100), where SD is your standard deviation for 100 hands, and N is the number of hands played. Note that SD is estimated by PokerTracker and given under session notes -> more info. Then your true win rate after N hands will lie within +/- the following multiples of SE from your PT win rate with the given confidences:

+/- 1*SE with confidence 68%
+/- 1.28*SE with confidence 80%
+/- 1.64*SE with confidence 90%
+/- 1.96*SE with confidence 95%

In general, for any confidence p, you can get the multiple x of x*SE from the Excel function =NORMSINV(0.5+p/2). Conversely, for any multiple x*SE, you can get the confidence p from the Excel function =2*NORMSDIST(x)-1.