How Many hands do you need to calculate winrate at a level?? This month I have 10k hands and 4.8bb/100. I know about variance a little but can I assume I am beating this level?
Printable View
How Many hands do you need to calculate winrate at a level?? This month I have 10k hands and 4.8bb/100. I know about variance a little but can I assume I am beating this level?
shaw
100k is decent, 500k is legit but by then the games evolved quite a bit that you can't be sure anymore. 10k is nothing.
is that for knowing if you're winning/losing or trying to get an accurate idea of your winrate?Quote:
Originally Posted by Alexos
i have no proof. but my best guess would be no. if you are still running at 4.8bb over 50k then i'd feel safer about it.Quote:
Originally Posted by gotigers1234
have a play around with this thing if you are interested in learning more about variance. ev++ Poker Tools :: Poker Variance Simulator
this winrate over this sample means that it is much more likely that you are a winning player than that you are a losing player at these stakes.
This.Quote:
Originally Posted by daven
It doesn't mean he's MUCH more likely to be anything.
95% confidence interval over 10k hands spans about 28bb
edit: Using 90bb/100 standard deviation
At the minstakes, by the time you've got enough hands for a good sample you've probably already moved up. Twice.
Having said that, what daven said.
The sample mean is positive. How can the % chance the true mean is positive be less than 50%?Quote:
Originally Posted by Icanhastreebet
x = (4.8 +/- 28) bb/100 :: (-23.2 < x < 32.8) bb/100
This contains more positive values than negative values, so what am I missing?
how did you come up with the formula mojo what do the numbers represent?Quote:
Originally Posted by MadMojoMonkey
I meant 40.6% confidence his winrate falls from 0-9.6bb/100 which is obviously pretty useless so I erased it from my post :).Quote:
Originally Posted by MadMojoMonkey
The numbers represent what your possible winrate is at 95% confidence, it's really hard to tell your true winrate over ANY sample, let alone one this small. I mean for example over 100k hands at 4.8bb/100 with std deviation 90bb/100 you can be 95% confident your winrate falls between 5.19bb/100 and -0.39bb/100 so even with this winrate over 100k hands you can't REALLY know you are a winner.Quote:
Originally Posted by gotigers1234
FYI: It takes about 200k hands at 90bb/100 std deviation to estimate your winrate with 95% confidence to + or - 2bb/100.
edit: It's supposed to be 10.38bb/100 and -.78bb/100(or read as BB/100 :)). My bad! I was in a rush when I made this post, the 2nd part is still correct though.
.Quote:
Originally Posted by Icanhastreebet