How many times?

[theriverwild]

Do you need to have gotten a hand before you can deterime if your win rate with that is accurate or not? For instance how many times do you need to get AA before your win rate for that hand is approximately the true win rate instead of one that may have a large part to do with variance and short term fluctuations?

[AaronBrown]

It depends on the hand, your style, the table and the limits. For example, it should only take one hand to figure out that your expected win on 72o is 0. That's a bit of an exaggeration, because you might hold it in the big blind and have people fold to you, but you don't need many observations. Something like JTs is much more chancy. You'll fold it sometimes, sometimes play and hit a flush or straight, sometimes play and miss the flop. When you do hit, or maybe pair on the board without overcards, you could have a big win or loss. How big depends on your style, the limits and the table.

Another problem is what "accurate" means. For example, suppose you found that your average win rate with JTs was $100, with a standard deviation of $1,000. Even after 100 hands, the mean would only be one standard deviation from zero. If the average win rate were $10 or $1, it would take an unrealistically large number of occurances to get the standard deviation significantly below the mean.

It makes more sense to look at statistics for groups of hands, say top pairs or high suited connectors. That gives you more observations to work with.