statistically significant hand #s?

[revlis87]

i posted this before but c ant find the thread.... so....

how many hands at a specific level do you need to know if your results are staistically siginificant? Thanks!

[joop]

100k is a start. 200k is better.

[revlis87]

and at 4 tables of NL how many hhands an hour would you get? im asking because i just updated my poker tracker after months and it only showed me 30 k hands... so obviously im missing an alias or something if i play 4 tables every day for several hours.

[d10]

This depends on what statistic you're looking at. It also depends on your definition of statistically significant. You'll never reach 100% confidence with any sample size.

[pzhon]

[ QUOTE ]

how many hands at a specific level do you need to know if your results are staistically siginificant?

[/ QUOTE ]
This is often asked, and people often give a number. This is not a good answer.

The standard deviation of your win rate in BB/100 after n*100 hands is about 15/squareroot(n) for LHE. A rough 95% confidence interval is your observed rate +- 2 standard deviations.

If you to have significant evidence that you are a winning player, so that 0 is not in the 95% confidence interval, you can often conclude this after only about 10,000 hands. Sometimes it takes fewer, sometimes many more, particularly for a marginal winner.

If you want to have a tight estimate on your win rate, you need many more hands. For the confidence interval to be the observed rate +- 1 BB/100, you need the standard deviation to be 0.5, so you need to play about 100*(15/0.5)^2 = 90,000 hands.

For NLHE, the standard deviation depends on your playing style. The 15 BB needs to be replaced by some number from 30-70 PTBB (60-140 big blinds).