| |
BGonline.org Forums
A better approch? [LONG]
Posted By: Maik Stiebler In Response To: A better approch? [LONG] (MaX)
Date: Saturday, 7 November 2009, at 5:52 p.m.
But an alternative could be to just use the CI approach to stop the rollout (similar to the current jsd stop criteria) and then compute the MonteCarlo estimate a posteriori (just once, at the end). if you're not satisfied with the resulting prob of the top player being best, just extend the rollout.
I like this a lot, as it would not slow down rollouts and still give all the interesting information.
FWIW, I experimented with another example where MinCI is misleading: You roll out N+1 move choices. All moves have the same standard deviation, and N of them have the same mean, while one of them has a higher mean by 2.5 SD's (i.e. 2.5/sqrt(2.)=1.768 JSD's). The PMC that that play is indeed best was 96.222%,93.235%,90.754%,88.685%,86.510%,85.084% for N=1,2,3,4,5,6 respectively.
| |
BGonline.org Forums is maintained by Stick with WebBBS 5.12.