Open Side Menu Go to the Top
Register
Sample size for winrate, +/- 0.3 BB/100 Sample size for winrate, +/- 0.3 BB/100

03-02-2012 , 08:19 PM
What kind of sample size do we need to conclude a winrate for 6 max, within around 0.3 BB/100?
03-02-2012 , 08:36 PM
It depends on how much confidence you want to have in that win rate. For 95% confidence, I think you need over a million hands. Since it's almost impossible to play a million hands under constant conditions, it's pretty much impossible to have that much confidence in your win rate.

I kind of hope I dropped a zero somewhere, but the moral of the story is not to worry about your exact win rate, with a sub-moral of make sure you're winning like twice what you need to live off of if you're playing for a living.
03-02-2012 , 09:22 PM
www.evplusplus.com
03-03-2012 , 07:58 AM
Quote:
Originally Posted by BigBadBabar
^

BOOM!
03-03-2012 , 01:50 PM
Does it matter that it's in bb/100, not BB? Is 40 an ok standard deviation? And won't it take more hands for FL than NL?
03-03-2012 , 04:02 PM
doesnt matter if its bb/100 , just enter the wr and standard deviation.

Standard eviation for 6max is about 18BB/100
03-04-2012 , 08:00 AM
Quote:
Originally Posted by pg_780
Does it matter that it's in bb/100, not BB? Is 40 an ok standard deviation? And won't it take more hands for FL than NL?
doesnt matter as long as u are consistent about using the same for SD and winrate, as the ratio winrate to SD is what determines how much u will swing.

I tend to think that for 500k stretches u can tell ur winrate with ok confidence within +/- 0.5BB.
Ofc "ok confidence" is a very relative measure but its something like 90%+ certain that u are within this range.

The kind of samples u need for added accuracy just takes too long for me too reach anyway, that too much will change in the meantime
03-16-2012 , 12:26 AM
Quote:
Originally Posted by GiantBuddha
It depends on how much confidence you want to have in that win rate. For 95% confidence, I think you need over a million hands. Since it's almost impossible to play a million hands under constant conditions, it's pretty much impossible to have that much confidence in your win rate.

I kind of hope I dropped a zero somewhere, but the moral of the story is not to worry about your exact win rate, with a sub-moral of make sure you're winning like twice what you need to live off of if you're playing for a living.
Nope you didn't drop anything. +/- 0.3BB/100 needs 1,000,000 hands for a 90% confidence interval. And I totally agree that game conditions are too variable to have a reliable 1M hand sample unless you can gain those hands in ~6 months or less.

      
m