How long before pokertracker stats narrow?

[Talex]

Hey Folks,

So BB/100 info takes a huge number of hands to be valid, and things like SD narrow more quickly. What about stats like VPIP, preflop raise percentage etc? Are there guidelines to figure how long it would take until those are reasonably close to your real numbers? I'm not enough of a statistician to even know where to begin to figure a confidence level on these sorts of numbers.

-Tim

Note: I assume this has been asked before, but I can't for the life of me figure a search that returns anything useful.

[Blindfolk]

Most people say 10-15k hands will let you know where you roughly stand. For it to be truly accurate, it would have to be in the hundred of thousands but if you are winning with 15k hands, you're a winning player at that limit.

[rivered]

As far as VPIP and the like, it seems to get steady very quickly. Even after a thousand hands it seems to be pretty close to my long term stats.

[Blarg]

VPIP and PFR zero in pretty quickly, but things like BB/hr and BB/100 take many tens of thousands of hands to be at all reliable.

That can be deceptive, because it's actually not that hard to go on a 20k winning or losing or break-even streak and think it's steady enough and a big enough sample to be the last word on how you play. Yet after months of collecting stats, give it a week later and you might be on a very different kind of run and quickly find your win rate stats changing remarkably. For win rate, I've still noticed marked fluctuations at over 50k hands.

Confusing things even more is that you tend to get better over time if you're a reasonably alert person who really tries to improve. So whatever stats you have are subject to change because you've gotten better, or change because you've tried new strategies, maybe very good ones, and screwed them up while getting used to them. Poker stats are very volatile things, and for some of them you really need big numbers.