Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Self-Play Experiments

Author: Christophe Theron

Date: 15:49:48 07/31/02

Go up one level in this thread


On July 31, 2002 at 11:48:38, Mike Byrne wrote:

>On July 31, 2002 at 06:19:08, GuyHaworth wrote:
>
>>
>>Ernst Heinz reviewed all past self-play experiments, with one-ply difference in
>>search-depth, in chess and other games.
>>
>>He also performed major self-play experiments:
>>
>>a) for his Ph.D.
>>b) afterwards, at M.I.T.
>>c) since - unpublished.
>>
>>ICGA_J v24.2 (2001) and ICCA_J v21.4 (1998) cover this.
>>See also:
>>    Advances in Computer Games, 9 - published by the Univ. of Maastricht.
>>    Scalable Search in Computer Chess, Heinz, Vieweg Verlag ISBN 3-5280-5732-7.
>>
>>Heinz's objective was to show, with a high level of statistical confidence, that
>>the 1-ply advantage in search-depth delivered decreasing advantage as
>>shared-depth increased.
>>
>>I re-analysed his stats in:
>>  7th Computer Olympiad Computer Games Workshop, Proceedings
>>
>>and it's clear that Heinz achieved his objective.
>>
>>Schumann has recently published similar results coming to the same conclusions.
>>
>>Caveat:  the experiments need to be carefully set-up to:
>>
>>a)  be repeatable by others
>>b)  ensure that they do not import factors which 'colour' the results
>>
>>g
>
>That's common knowledge - you don't need a PhD to figure that one out - just
>play around with chess programs ;>)



But you need a PhD to PROVE it...



    Christophe



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.