Well, "I back and I'm proud"

[ghostface]

[ QUOTE ]
that ratso person is really a looser

[/ QUOTE ]

[pudley4]

[ QUOTE ]

JEFFA…You are correct to a point. Please explain standard deviation to us.

[/ QUOTE ]

Your reading comprehension sucks.

[ratso]

Jeff A:
I was serious about explaining the "standard deviation" when it pertains to poker. I do not believe that calculating the standard deviation for poker is trivial. Do you calculate this number simply as one would by using the Excel function on a column of wins and losses? I do not know, really. What is more important? Is variance or standrd error more meaningful? Does one have to use a weighting? How can one account for differences in different levels (i.e. 10/20, 20/40, NL 1-2 or 5-10)? I guess they should all be separated. If the player is good, then one would expect more wins than losses and naturally more $$$ which would suggest non-normally distributed data. How does one handle that? I do not believe a mean +/- the stdev is technically correct. Should the data be transformed? These are tough questions for which I have no answers. I have not read any sources that address these questions. I (and others, I suspect) would love to see any reference about calculating the most meaningful statistic that would give players a handle on their variance or swings. I suppose Mason has considered this. Perhaps he has an answer. If you could provide some info I would be grateful. Seriously, an apologetic and humane Ratso.

[Cosimo]

re: calculating standard deviation.

There used to be an Essays page here, but I can't find it. Mason's essay on calculating standard deviation was reprinted there. If you have Gambling Theory and Other Topics, I believe it's in there.

You will have different SDs at different levels, limit vs NL, full vs SH vs HU, different sites, and different games (HE vs Omaha vs Stud etc). Generally you can use 15 BB as your standard deviation.

If you calculate your EV and standard deviation, you will be able to calculate "for four hours at live play 10/20, I should win $56, with a standard deviation of $600." I believe those with huge databases that have charted the data are, indeed, getting Gaussian distributions of their session results. Poker winnings are discrete, of course, so the shorter the session and the fewer the sessions, the less proper such a graph will look.

SD is calculated as RMS of the differences between individual session results and the expected result from that session. "Expected result" here is just EV multiplied by time at the table. Yes, it is easy to calculate ...

... but confidence is very low. You have said that you've played 12 hours a day, two days a week, for three years. I'm guessing that is about 80,000 - 150,000 hands, depending on how precise and consistent you've been. That number of hands is about enough to have a confident assesment of your EV -- provided that your skills and the game didn't change during that time, both of which probably did change signifigantly.

So, again, precise values for standard deviation and EV are not really possible, and chasing them is not too useful.

You should find more on this topic by searching and/or posting in the Probability forum.

[ratso]

Thank you. Excellent explanation. I understand. I actually had a discussion with MM 5-6 yrs ago about just this. Seems there is a term, variation, that is often mistaken for stdev. I do know the statistical difference. I do have the Poker Theory book(s) and will look it over. I calculate my SD simply using Excel after sorting by game, and I find my SD considerably larger than 15 BB in a 10/20 HE game, yet my earnings approach 2 big bets/hr net. Surprisingly (at least to me)in low-mid limit stud my SD is quite small, and my wins are over 3 big bets/hr net. Thanks again.