Author: Uri Blass
Date: 02:24:58 08/07/04
Go up one level in this thread
On August 06, 2004 at 22:53:03, Robert Hyatt wrote: >On August 05, 2004 at 23:33:59, Mike S. wrote: > >>On August 05, 2004 at 23:02:40, Bruce Humphrey wrote: >> >>>1) Use tablebases. Not sure how it would affect ProDeo, probably it would help, >>>at least 5%? >> >>Never! You mean when it scores i.e. 55.0% against an opposition, it would score >>60% against the same, with tbs.? No... I guess it would be more like 55.5% or >>maybe 56%. I would not expect more. Half a point once in a while... > > > >Crafty just played Pro Deo on ICC, 6 games. the first 5 were normal games, 1 >win for Crafty, 4 draws. The last game was a dead drawn KRN (crafty) vs KRPP >(Pro Deo). Crafty won due to endgame tables. They _do_ make a difference from >time to time, and not just once per 100 games or so... Dependent on the opponents that you play and on the knowledge of the program. It may be interesting to have a test suite based on positions that programs blundered because of not using tablebases when the target is not to blunder. I suspect that some knowledge in evaluation may prevent most of the blunders and also may be productive in non tablebases positions. 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.