Author: Robert Hyatt
Date: 10:06:15 10/30/03
Go up one level in this thread
On October 30, 2003 at 12:12:51, jefkaan wrote: >On October 30, 2003 at 11:35:05, Robert Hyatt wrote: > > >On October 30, 2003 at 11:35:05, Robert Hyatt wrote: > >>Comparing scores between two engines is just like comparing search >>depths, or anything else. > >hello mr Hyatt, well thx for your reply, >but actually i meant at the *same* (or similar) >search depth, or similar time controls/cpu strength; >at higher depths, the score of one engine usually >seems to 'converge' (although this usually is an illusion >as the search after 14 ply or so seems to go 'slower' ..) > I was answering for _all_ cases. IE chess tiger and fritz, at the same depth. The evaluations might be significantly different. The absolute value of the evaluation doesn't necessarily mean anything. One program could have +1.0 in a winning position, another might have +5.0 in the same position.. IE the CSTal program often had inflated scores due to its speculative style of play... >So what i really meant was: how big can be the >differences in positional evaluations between >different top-engines ? And the reason i mention >the letter t*o*p* is because otherwise an imbalanced >evaluation can cause rating points (and ofcourse forget >the scaling factors, in the end things come down to material >evaluations like a queen is approx 8 pawns and in the end >mate=mate). Remember the program Mchess (version 7. was quite >good, interesting opening book, human style of play etc.) ? >When looking at this program I've really seen some differences >with other engines more than 1.5 pawn value, but in the >end after proper analysis Mchess usually wasnt right >(Crafty on the other hand could still benefit from >some more emphasis on positional >vs material evaluation i suspect) > >best regards, >jef
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.