Quote:
Originally Posted by scylla
If you have dealt-based rakeback then for simulation purposes you might as well asume you have no rakeback at all I imagine.
Does the stoxev rakeback calculation take much processor power?
**************
I have a question about Variance, Standard Deviation, and Sample Size. It's been years since I've taken statistics classes and my brain is rusty.
As an example, imagine that I am running a simulation where my range of opening hands is quite wide, as is my opponent's range.
I would like my simulation to tell me whether each of my possible holdings is +/- EV.
Furthermore, I am testing an action which is results in a huge variance. For this example, let's say each player is betting/calling allin for $1000 into a pot with no money.
I run my simulation and get back the results. My EV is 40, the standard deviation is 15, and variance is a whopping $950.
Of course, if I have run this simulation 100,000x1,000 times, the result will be fairly accurate. But if I run it only 10x1,000 times, it is certainly unreliable, especially considering that this reflects a large variety of holdings.
Is there an easy way to estimate the sample size I would need (in other words, the number of runs) to obtain a decently accurate result for these scenarios that have wild variance?
Or do I need to use the statistical sample size/variance/std. dev. formula to have any hope? This formula encounters problems since I would like to use the results from each of my possible holdings. (But of course the same problems would occur while estimating) The only real solution to this, I think, would be to run each hand combo separately, which would be too hard.
Any input into this?