Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Help ! Fritz 5 engine bug ????

Author: blass uri

Date: 05:28:10 11/02/98

Go up one level in this thread



On November 02, 1998 at 06:35:46, Bernard Nahrmann wrote:

>I have a problem with Fritz 5 and this position :
>
>8/8/8/8/2k3p1/P3P1Pp/7K/8 w - - 0 1
>
>I let Fritz 5 analyse this position in analyse mode !
>Showing one variation Fritz finds soon Kc3 = with value 0.0 !
>Showing two variations Fritz displays Kc3 and Kb3 with value 0.0 for eternity,
>but Kb3 is a clear loss. Playing Kb3 Fritz sees this after a few seconds.
>Showing three variations Fritz shows Kc5, Kc3 and Kb3 each with value 0.0. For
>Kc3 and Kc5 this is o.k., but again Kb3 ??
>Showing four variations Fritz shows, that only Kc3 and Kc5 are =, Kb3 soon
>disappears !!!???
>
>I tried this with the current Fritz 5 engine with different hash sizes under
>Fritz5.16, Junior 5 and Chessbase 7.0. I also tried it on different machines and
>Win95 and NT 4.0.

I did not try this position but I know from experience that fritz5 has a bug in
the x variation mode.

The bug is that the score of the move that fritz5 "thought" to be best cannot be
lower than the score of the second best move.

Example:suppose fritz5 "thinks" that  the score of the best move (Kb3) at depth
8  is +2 and the score of the second best move (Kc3) at depth 8 is 0
then it cannot discover at depth 9 that the score of kb3 is lower than 0.

If the score of Kc3 is never lower than 0 at depth 10,11,12,... then fritz5
never discovers that the score of Kb3 is lower than 0.

Uri



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.