Poker-AI.org

hello, right now I am working on the problem how to estimate for example the population distribution of vpip. The problem I see is that if I just take the distribution from a sample of lets say 100000 hands I will get a distribution with probably the same mean as the population but a higher variance. Now one thing I could do to lessen that bias of variance is to only include players that have played for instance more than 200 hands. however this itself might introduce a new bias.
I have read around a lot about statistics during the last days, but just didn't come up with a consistent solution. So just wondering if there even is a well known better solution.