Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: ?! But this is useful information

Author: Mike S.

Date: 15:13:28 11/21/00

Go up one level in this thread


Because, such imperfections can (and most probably will, if the number of games
is very high) affect the rating of a program in a way, that it is underrated
because it draws "won" games. Some testers, but very few, may look into all such
test games and see that a program may be stronger than it's result indicates -
But what should they do? The score must not be manipulated, and most often the
pure result figures are the main information that is transmitted and consumed.

After all, there is a rating list thinking in computer chess. Therefore it is
most important, that a program makes the full point from a won game. Otherwise,
possible customers are misinformed unavoidably, at least to some degree.

So I'm sure programmers are always interested in such info, and it is valuable
for them. Btw. there have been cases IIRC where such, or similar, endgames are
implanted and tested, but wouldn't work due to certain special unexpected
conditions (I remember that a Fidelity computer could solve a tactical? position
when it had the original knights from the starting position, but not with
knights which were inserted by a board setup, for example.).

Regards,
M.Scheidl



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.