I've used statistical methods to estimate that my "true" live win rate is somewhere between $18 and $51 per hour, with 95% confidence.
I've logged 1037 hours of live poker, and I track the length of each session. Using this data, I have a script which calculates the sample standard deviation of my win/loss per hour, and which then calculates a confidence interval for win rate using the recipe I found here:
http://answers.yahoo.com/question/in...3162427AALf9tS
The interval I found is a lot smaller than I thought it would be, which I think indicates that a mere 1000 hours of live play CAN potentially be enough to get a very rough idea of your "true" win rate, with a few caveats...
I have a very tight, very passive style, which I assume means that I suffer from significantly less variance than most poker players. A more aggressive player probably needs significantly more hours before they can achieve statistical significance.
In fact, it seems that the number of hours required to achieve a given confidence interval is proportional to the SQUARE of the standard deviation. So, for example, if a given player was ten times as aggressive as me, they would presumably have an hourly standard deviation that was ten times larger than mine, and they would therefore need to log ONE HUNDRED TIMES as many hours as I've logged before they could be as sure as I am of their win-rate, even if our win-rates were exactly the same.
And of course, there are all the normal caveats about the games always changing / getting harder / sampling bias / logging errors, etc, which could easily make my analysis meaningless.
Apologies if anything I've said here is incorrect. I just worked out the math/code today, following some random recipe I found on yahoo answers, so everything I've said could easily be wrong (also, I got a D in probability/statistics in college ;P).
Edit: My "standard deviation per hour" is $271, in case anyone wants to compare numbers. I see that some other people have already posted theirs in an iphone screenshot. If you want, you can PM me a) your hourly rate, b) your stddev/hour and c) the number of hours you've logged, and I can tell you what your 95% confidence interval is.
Last edited by BoredAtheist; 01-06-2012 at 12:20 AM.