Author: William Penn
Date: 15:55:18 12/10/00
Go up one level in this thread
Here's an example of a position that flip-flops the sign attached to the score when it goes above/below the critical 1.00 value: -- -- BR BR -- -- BK -- BP BB -- -- -- -- BP BP -- -- -- -- BP -- -- -- -- -- -- -- WP BP -- -- BQ -- -- -- -- -- -- -- WP -- -- -- WQ WP WP -- -- WP -- -- -- -- -- WP WR -- -- -- WK WB -- WR (Black to move) When the score number is bigger than 1.00, the minus sign is attached. When the score number is 0.99 or smaller, the minus sign is dropped (probably incorrectly so). This evaluation is with the default Chessmaster personality and engine. The same thing happens with custom defined personalities. Time Depth Score Positions Moves 0:00 5 -1.21 26737 Qc2 0:02 6 -1.14 90450 Qc2 0:07 7 0.90 269558 Qc2 0:25 8 -1.08 871667 Qc2 1:10 9 0.99 2628902 Qc2 5:31 10 0.99 11104747 Qc2 23:30 11 0.80 48168538 Qc2 WP On December 10, 2000 at 17:55:12, William Penn wrote: >Please take a close look. I believe that James T. Walker has put his finger on >the evaluation problem as expressed in scores: There is apparently a bug when >scores are in the range of +1 to -1 which fails to put a minus sign on negative >scores. The minus sign is only affixed to the score when it is less than -1. >WP > > >On December 10, 2000 at 17:11:19, James T. Walker wrote: > >>On December 10, 2000 at 16:58:07, William Penn wrote: >> >>>To Jeremiah & James... >>> >>>Thanks to Jeremiah Penery and James T. Walker for your replies and suggestions. >>> >>>For some unknown reason, I am unable to post a reply in the original thread >>>below. I just get a message saying the thread doesn't exist, hasn't been posted >>>yet, may have been deleted...etc. I may have a corrupted cookie, or something. >>> >>>There is a dearth of documentation about CM8000, and nothing in the instructions >>>about such things. We are apparently dependent on John Merlino to come along and >>>fill in the gaps, insofar as he knows the solutions/answers. I've heard from >>>him before on this subject, and am not sure that he knows the answer. It may >>>require the chess engine's author to take a look at it in a very patient way, if >>>he is willing, but that could be a problematical proposition too!? >>> >>>Bottom line: I can't be confident in CM8000's scores generated when evaluating >>>positions, and that's pretty fundamental. >>>WP >><snip> >> >>Hello William, >>I have finally been able to understand the problem. You are right about >>ChessMaster 8000 having a problem with the scores! All scores below 1.00 are >>positive even when black is winning. As soon as the score goes over 1.00 for >>black the scores show the minus sign. This is a definite error in the GUI. I >>don't think the engine has a problem since it is playing so well (except for not >>using it's time). I think it's just a bug in the scores below -1.00. I just >>finished a game in which the score was around 0.97 and I'm sure black was ahead. >> The scores then bounced from 0.97 to -1.01 back to 0.91. I'm sure the position >>was not that wild. It's just that all scores below 1.00 are positive. I have >>noticed this before but could not put my finger on the problem. I noticed that >>ChessMaster seemed to always think white was winning untill something drove the >>score past -1.00. Then it shows scores like -1.21 indicating black is winning. >>This is definitely a problem for John to pass along. >>Jim
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.