Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A theory of ratings drift for the SSDF

Author: Ricardo Gibert

Date: 22:27:06 04/09/02

Go up one level in this thread


On April 09, 2002 at 23:02:38, James T. Walker wrote:

>"The most recent games have a far greater impact on the rating than the earlier
>games. Do not be impressed by the large number 1200. It does not matter. 1200 is
>the same 24 (or whatever it is nowadays to get an established rating)."
>
>I don't believe this is the case.  I believe the Elo rating is computed by SSDF
>the same way Chessbase/Fritz GUI does it for an entire database at one time.  If
>this is true then the last 24 games have no more effect than the first 24 games.
> Someone from the SSDF can correct me if I'm wrong.  If they did it the way you
>suggest the ratings would fluctuate much more from month to month.
>Jim

This is plausible, since the a program (a particular version) does improve over
time. At least if learning is disabled. After a few hundred games, the rating
will not change very much anymore and the purpose of pairing it would be to
produce ratings for newer programs using the older programs as a benchmark.



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.