[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Bug-gnubg] rollout questions

From: Christian Anthon
Subject: [Bug-gnubg] rollout questions
Date: Tue, 25 Apr 2006 10:51:05 +0200
User-agent: Mutt/1.5.10i

Hi all,

a couple of rollout questions:

1) Does anybody use the stop-on-small-std option. The std.
deviations/result ratio does appear to be a very good criterion.

2) What is the difference between to two-stop-on-high-jsd options. The
explanations given in the gui makes it sound like they are equal to me.

3) Using stop-on-high-jsd doesn't work very well on take-drop
decisions as these are usually clear doubles.

4) are the following two evaluations equivalent (the results are very
close by not quite equal)

a) 0ply play and cube, truncate at ply 4 using 0ply, rollout 36^4
times, quasi-random dice, no variance reduction
b) 4ply cubeful evaluation

If that is the case, we could get very fast and quite accurate 4ply
results by using 0 ply rollouts truncated at ply 4 with variance
reduction and say 216 games since the variance reduction works very well
on such few trials.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]