Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: an idea to test positional understanding of chess programs.

Author: Robert Hyatt

Date: 14:57:01 05/24/99

Go up one level in this thread


On May 24, 1999 at 16:04:36, blass uri wrote:

>The idea is to do a match between chess programs at tournament time control when
>both sides have the right to take back moves(they will use it  if they discover
>that they do a tactical mistake).
>
>Every program lose 5 minutes in the time  when it take back a move
>and the opponent earns 5 minutes.
>
>Programmers should change their program to consider take back but I do not think
>that it is a big change.
>
>It is logical to tell the program to calculate if to take back a move only when
>there is a big change in the evaluation against it and it can be productive only
>if the mistake was a tactical mistake(otherwise it should take back many moves
>to avoid losing and it will lose on time because it is losing 5 minutes for
>every take back).
>
>What is your opinion about it?
>
>Uri

This is already how I 'debug' Crafty...  But I am not sure how I would make
it 'play' such a game.  IE if it makes move 31. XXX, and then the score drops
was the problem with move 31. XXX, or was it with move 30. YYY?  IE it might
take back 31. XXX and try 31. YYY only to see that drop too...  and then try
another move 31 and so forth.

It would be 'interesting' to make it work...  Because this is another form
of 'learning'.... ie figuring out where it screwed up and taking back _that_
move rather than taking back a random move.



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.