When to quit?

[Atropos]

Hey folks,
this question has probably already been asked but I dont know what exactly I could search for...
I play the NL 6-Max Games at Party, 4 tables at a time. The stack size there is 100 Big Blinds. My Standard Deviation (insufficient sample size) over 100 hands is 50.3 BIG BETS meaning 100.6 Big Blinds / 100 hands. What does this number tell me?
Is it possible to use this number and get a number of lost buyins that is unlikely enough, that it would be better for me to quit, since I'm a player that cant control tilt too good? I want a number like "lose 2 buyins - quit for the day", but this number may not be too small because I could not play too often /images/graemlins/smile.gif and this number may not be too big because I could have been tilting for a long time already...

[Orpheus]

[ QUOTE ]
My Standard Deviation (insufficient sample size) over 100 hands is 50.3 BIG BETS meaning 100.6 Big Blinds / 100 hands. What does this number tell me?

[/ QUOTE ]
Your SD over 100 hands? Two hours of play?

I'm not being snide when I say: it only tells you that your sample size is insufficient. The size of your reported SD is TOO high to reflect your play accurately. If you mixed my life stats with a WSOP champion, you still wouldn't get an SD>50 BB/100 (and I'm not very good). With an SD that high, you could literally be a rainbow trout that has been trained to fold bad opening hands (-30 BB/100 longterm) or a superior alien capable of +30 BB/100 for centuries -- and the trout could be sucking the alien dry over 100 hands, with a moderately good run of cards.

Let's I flip a perfect coin twice, I will either get SD=0, (both heads or both tails) or SD=0.5 (one head, one tail). Both results tell me next to nothing about the underlying phenomenon. In fact, since I'm equally likely to get SD=0 and SD=0.5, I could be flipping a two-headed coin, and I wouldn't know it.

Considering that I haven't studied the SD of a binomial distribution since Jimmy Carter was President (I was precocious), I may have the statistical probabilities of a coin-flip slightly wrong, but I believe...

The mean number of H for a coin (biased or unbiased) is p*N (p = probability of Heads; N = # of flips)
The standard deviation number of heads is sqrt(p*(1-p)*N)

After 10 flips, you'd expect P=0.5, but out of the 1024 equally likely series outcomes, only 252 (less than a quarter) have exactly 5 heads. Six H heads (observed p=0.6) or 4H (observed p=0.4), are almost as likely, with 210 outcomes each. Series with two (or eight) heads are surprisingly common (4.4% each). Even a single head or tail is only a hair under a 1% probability. The only truly unlikely outcome is all heads or all tails (<0.1% each).

Aith a sample size of 10 flips, you can't even get a reliable idea of the mean of something as simple as a coin flip. The SD of ten coin-flips is 1.58, so you'd probably need 100 or more trials to be reasonably sure of getting a result near 0.50. In fact, your coin could be quite biased, and you still might not be able to tell with 90% reliability after 100 flips.

Can you see why 100 hands is insufficient to evaluate poker play, which has many more choices, and a far larger range of outcomes, compared to a coinflip 0 or 1?

Movies aside, it could take 1000s of throws of a 'normal' loaded die to be 95% (2 SD) sure it was loaded. Such dice still have to feel "normal" in the hands of an experienced croupier. Poker is much more complicated than craps or coins.

100 hands isn't enough to see all the *common* hand situations even once, much less, most of the important money -making and -losing situations. In 100 hands, you very rarely have as many 5 real "judgement calls". The overwhelming majority of hands are folded preflop. How much can they say about your quality of play? (Unless, of course, you play even the worst of hands, but then you'd have an SD of 50BB/100 with a hugely negative mean)

Aside from the game, your play can change from session to session (and frankly, it probably *should* be changing steadily over your first 10,000+ hands.

[Atropos]

Thx for your but I dont think your text applies here because there is one error ( or misunderstanding): By SD/100 hands I did not mean my Standard Deviation over only 100 hands but the average number Pokertracker gives me...