Author: Bob Durrett
Date: 06:44:05 12/21/02
Go up one level in this thread
On December 21, 2002 at 08:49:26, Uri Blass wrote: >On December 21, 2002 at 07:58:59, scott farrell wrote: > >>On December 21, 2002 at 01:48:41, Dana Turnmire wrote: >> >>Great thread. I am working on this sort of thing at the moment. I have been >>trying to make it play GM moves, with lots of sacs, and positional trickery. > >I think that it is a mistake to try to do the same as humans because I believe >that the assumption that GM's play better positional moves is wrong. > >There are also cases when there are a lot of positional moves that lead to >almost the same evaluation. Perhaps "positional" test positions, in test suites, should give several solutions with some solutions given more points than others. I vaguely recall a chess book titled something like "Point Count Chess." The author gave the reader the most points for finding the best move, but also awarded a few points for lesser solutions. Maybe that should apply to chess engines too. An engine that doesn't necessarily always find the very best move may still be able to play a decent game of positional chess. Test suites aimed at measuring a chess engine's prowess at positional chess perhaps should be multi-solution. Bob D. > >If trying to learn to play moves of strong players is productive then >I see no reason to try to play the moves of GM's and not to try to play the >moves of the ssdf leaders(for example tiger's moves afainst ruffian). > >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.