True winrate

[Derek in NYC]

Variance is both a bitch and a goddess. Assuming a pretty normal SSH standard deviation, at what hand count do you start feeling comfortable that Poketracker winrate (BB/100) is accurate? Directionally accurate?

[AliasMrJones]

This isn't hard to analyze yourself. Get a spreadhseet (openoffice.org will do fine and is free.) The formula for worst/best case with a given confidence level is:

(Win Rate * Hours Played) - (normsinv(Confidence %) * Standard Deviation * sqrt(Hours Played) (worst case)

to

(Win Rate * Hours Played) + (normsinv(Confidence %) * Standard Deviation * sqrt(Hours Played) (best case)

The mythical "really good" online player is 2 BB/Hr and 15 BB Standard Deviation. Plug the confidence interval of your choice and see where the best case and worst case converge to an absolute value you like.

At about 300 hours you're at +/- 2BB/hr with 99% confidence. (The break-even point. If you're ahead at that point you know you're a winning player.)

At 10,000 hours you're at +/- .35 BB/hr at 99% confidence. (That's assuming you're at 2 BB/hr and 15 SD after 10,000 hours.) at 60 hands per hour that's about 600,000 hands.

How accurate is "accurate enough" for you?