Poker Tracker Stats converge

[jayheaps]

quick question about PT stats when looking at opponents. After how many hand do you think an opponents stats, particularly VPIP and aggressiion will be within a few % of their "true" rate?

In other words after how many hands can opponent's stats be reliable? I am looking for more of a math answer than an intuitive one.

For me, i'll start to look at stats with 75 hands however i dont put too much weight on them until around 250.

[sweetjazz]

It converges faster for a tight player than a loose player. Someone who plays only 10% of hands before the flop will see his VPIP drop below 20% pretty quickly. Someone who plays 35% of hands could have a VPIP above 50% for a longer period of time.

Or at least I think this is right, but I am too tired to give a mathematical argument right now.

[VBCurtis]

For binomial distributions (VPIP reduces to fold or not fold), std deviation is a function of the product of the percentage of times the two choices are made. Thus, the reply was correct-- 10% VPIP times 90% fold is .09, while 35% VPIP times 65% fold is roughly .20, meaning the std Dev of the loose player is double the tight player.

Thus, it takes either double the hands or 4 times the hands (square root of std dev? I forget) to have an equally accurate read on the loose player for VPIP.
-Curtis