Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Scalable Search... Errata?

Author: GuyHaworth

Date: 01:43:59 07/02/02

Go up one level in this thread



MIT tell me that Ernst Heinz' email addresses are occasionally consulted,
although he's now somewhere in Germany I think.

You can get these from deacon@mit.edu if you wish.


I have left a message there for Ernst recently, as I believe that his
statistical analysis of the self-play work, particularly his own major
experiment which followed the book, does not provide the higher statistical
confidence figures which are available.

Admittedly, this makes little difference to previous small-scale experiments:
they ran too few games per match to make an impact.

However, for his 3,000-games/match experiment, the use of a win/draw/model and,
more importantly, some standard statistical stuff about independent variables,
does result in higher statistical confidence limits.

I am covering this briefly at the ICCA Computer Olympiad Workshop in Maastricht
next week.

g



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.