Author: Uri Blass
Date: 09:28:24 01/07/02
Go up one level in this thread
On January 07, 2002 at 11:50:08, David Rasmussen wrote: >On January 07, 2002 at 10:50:34, Severi Salminen wrote: > >> >>Of course in most of the cases, you don't need to worry about 9 queens getting >>involved in the same SEE calculation, but my king value is big and I use 32-bit >>ints to store the score so that's not a problem for me. But I think the values >>won't balance out (unless you meant material score, or something) in SEE > >I meant that I am using the same piece values for material score as I am using >in SEE. When I do SEE, I just use an int, so that should be fine, but as the >same values are used for material score, which I use in evaluation which >eventually ends up in hash entries, I want the material scores to not be too >big. I already use many bits (18) for storing the score, because in this version >of Chezzz, I am using millipawn values instead of centipawn values (I want to >experience myself that millipawn precision is not worth the effort before >switching back to centipawns). I believe that chess is a tactical game so centipawns are enough. I believe that if somebody decides to chenge the evaluation from centipawns to 1/50 pawns there will be less than 10 elo change in the rating of the program. It may be interesting and I think that it is easy to change the evaluation to be always even in order to check it. You only need to tell your computer to add 1 for the side with the advantage if the evaluation is odd. 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.