Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A test position for chess programs

Author: Uri Blass

Date: 12:16:23 12/10/00

Go up one level in this thread


On December 10, 2000 at 13:16:06, James T. Walker wrote:

>On December 10, 2000 at 11:47:15, Uri Blass wrote:
>
>>On December 10, 2000 at 11:35:10, Daniel Chancey wrote:
>>
>>>The personality, Chessmaster8000 (HB=16) found Rd5+ at 5:34 although the score
>>>increased from 0.47 to 0.55 (favors white).  The move stayed through 10 minutes.
>>>
>>>Castle2000
>>
>>Black played Rd5+ so if the score increased from 0.47 to 0.55 then the score has
>>to favour black.
>>
>>Uri
>
>Hello Uri,
>In ChessMaster a + score is good for white so an increasing + score is even
>better for white.  I gave the position to Rebel Century 3.0 on my Athlon 900 and
>it finds Rd5+ after only 16 seconds but changes to Rb4 at 28 seconds.  Then it
>switches back to Rd5+ after 1:06 and keeps it past 6:45 and 12 plies with a
>score of 0.91.  I guess it will stick with it.
>Jim

Hello Jim,

I do not have chessmaster8000 so I do not know the meaning of the scores of it
and I used simple logic.

Simple logic says that if Chessmaster changed its mind about the best move the
score should be better for the side to move(black).
Simple logic says that if the score was +0.47 for white then the score cannot be
changed to 0.55 pawns for white when chessmaster suggest another move for black.

I can explain behaviour like this only by hiding part of the information or
lying about scores.

Chessmaster needs to see a fail low for the move with evaluation of +0.47 for
white.
If chessmaster does not tell that the customer that it found a fail low for the
move with evaluation +0.47 for white then it is hiding part of the information
about the scores.


Another possibility is that there are cases when chessmaster fails low without
solving it and that the score of chessmaster is only a bound for the real score.

I can understand if the score is changed from  at least 0.47 pawns for white to
exactly 0.55 pawns for white(assuming that chessmaster decided not to solve the
fail low problem) but in this case the score should be >=0.47 for white that
should be translated to a number that is bigger than 0.55.

If chessmaster says 0.47 for white and not >=0.47 pawns for white in this case
then it is lying about the score.

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.