BGonline.org Forums

Variance Reduction

Posted By: rambiz
Date: Monday, 21 July 2014, at 3:01 p.m.

In Response To: Variance Reduction (Bob Koca)

TYVM and my best wishes for you and your better half. May God bless your marriage.
Yes I need to know how much VR helps or more precisely I need to know the variance of the variance-reduced estimate for each single trial.
Meanwhile I was thinking of the following trick: Given we had access to the variance-reduced-estimates (an estimate of the winning chances and not a zero or one, as you kindly elaborated upon earlier) after each trial, calculate the sample variance after n trails, using the obtained data. Then use this estimated variance along with the normal distribution to check if we are already inside the desired confidence interval. The problem is of course, that we will have only an estimated variance. So actually one needs to somehow know how accurate this estimate is, and then maybe given some independence assumptions multiply 2 numbers to get the real probability of say equity being within certain confidence interval.

You sugest: If you want the confidence interval to be a certain size you just need to wait until it gets there.
Since I can't reproduce the reported numbers,I have no faith in what the bot says, in the first place.

Post Response

Subject:
Message: