Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Endgame tablebases and draw by repetition : a bug in Crafty 16.5T (Fritz

Author: Robert Hyatt

Date: 12:32:13 07/12/99

Go up one level in this thread


On July 12, 1999 at 15:21:18, Sylvain Renard wrote:

>On July 12, 1999 at 10:16:11, Robert Hyatt wrote:
>
>>On July 12, 1999 at 03:46:02, Sylvain Renard wrote:
>>
>>>Using Crafty 16.5T (Fritz interface) with endgame tablebases,
>>>I have noticed this bug (easy for me, my program has the same!!) :
>>>
>>>Start position:
>>>
>>>4k3/8/6P1/4K1P1/8/8/8/8 w - - 0 1
>>>
>>>White: Ke5, pawns g5 & g6
>>>Black: Ke8
>>>white to move (machine = black, white = weak player!).
>>>1.Kd6?? (there are several wining moves!) 1...Kf8= 2.Ke5 Ke8?? leading
>>>to the start position, so Crafty 16.5T (Fritz interface) evaluation is 0.00 ,
>>>because of the repetition of the position!
>>>  This bug is not present in "normal" Crafty 16.11 (Winboard interface) so I
>>>wanted to ask Mr Hyatt his explanation.
>>>  Thank you and best regards,
>>>    Sylvain Renard
>>
>>
>>There doesn't appear to be anything wrong.  Crafty considers two repetitions a
>>draw, as it always has.  In that regard, the above is a draw as far as it is
>>concerned.
>
>   A program should never play a losing move in this position!
>Now the opponent can play 3.Ke6, 3.Kf5 or even 3.g7 and win.
>I still consider that this is a bug!


In a way it is.  But then again, the human screwed up the first time he made
a move too.  Fixing it is easy.  But it hurts performance.  And it makes it
much harder to find some cute draws that happen.  I personally accept the
problem as it generally is a non-issue...



This page took 2.68 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.