Re: Our PS winrates
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
what is long?
people say: play 10k hands than post you stats her for review
others say: play 100k to eliminate swings...
I will post some datamined charts and i would really like to discuss:
"if you can beat the game"
[/ QUOTE ]
You need 30,000 hands at a level before knowing your general winrate, and 100,000 hands at a level before knowing your specific winrate.
[/ QUOTE ]
Such a crock.
Assume 18BB/100 standard deviation, which I understand is normal if you work it out right. Then over a 100,000 hand sample your standard error is
18/sqrt(1000) = 0.57
so with 95% confidence you know your winrate to within plus/minus 1.14BB/100
That is not particularly specific.
[/ QUOTE ]
Guess it depends on what you mean by specific.
BTW - if your winrate is 3BB/100 after 100,000 hands, your equation means that your true winrate is 2.43BB to 3.57BB. (Not 1.86BB to 4.14BB).
To me, that's pretty specific.
[/ QUOTE ]
That would be a one standard deviation confidence interval, i.e. about 70% confidence. For 95% confidence it's 1.86 to 4.14, i.e. two SDs either way. And remember that 5% is not exactly a small chance we're wrong, so let's go for three SDs, 99.7% confidence: 1.29 to 4.71BB/100
Is that specific?
And given that most people are looking at 1.5BB/100 winrates max, we're talking about an interval of -0.21 to 3.21BB/100 after 100,000 hands. Not specific at all in my opinion.
Guy.
|