Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Help with strange tablebase output - Crafty-18.6 & Fritz6

Author: Dieter Buerssner

Date: 07:55:18 03/17/01

Go up one level in this thread


On March 17, 2001 at 06:40:59, Mike S. wrote:

>It's not relevant in this case, but Fritz 6 has problems with tablebases &
>repetition. This is a known issue. I remember another example which was
>discussed in the CSS Forum (with Mathias Feist and others) some time ago, where
>Fritz blundered in a tb situation by wrongly repeating a position. This was out
>of a "natural" move sequence, not after a deviation like 1.Rg1+ here.

Perhaps, we remember the same discussion in the CSS forum. But I remember it
slightly different. In that position, a repetition of position could be forced.
Nevertheless the position was won (by using another plan after the repetition).
It looks very similar to case in this discussion to me.

It is a design question for the program author, how to handle this. Many chess
programs seem to give a draw score after the same position is seen twice. This
will work well in almost all games. But this can give surprising results,
especially when analyzing or forcing moves. Some programs, including Crafty,
Yace, and probably Gandalf will not evaluate a position that is only seen the
second time as draw (at least not, when the repetition is close to the root
position). This in principle should make the following possible: After some
moves, the engine sees, that the opponent can force a repetition. However the
engine thinks, it has the better position. So it might look for another plan to
win. But it also can yield in a loss. I believe the games Gromit-Yace and
Fritz-Ganddalf are examples for this :-(

Regards,
Dieter





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.