Author: Mike S.
Date: 05:38:55 06/26/00
Go up one level in this thread
On June 26, 2000 at 02:39:43, blass uri wrote: >(...) >It is possible to do a test suite based on positions that humans understand but >it is only part of the practical positions that program have to play when they >do not see a way to win material. > >Uri Yes, it is very difficult to find good and reliable test positions for the positional abilities. However, from 10 such more or less qualified positions which I use, I get the best results for Chess System Tal II (but it isn't on the SSDF list), followed by Socrates 3.0 (!), Comet (I think B02 version; I don't have my detailed results available at the moment) and Shredder 2 or 3. Most of the tests need some kind of sacrifice to gain a big positional advantage, for example the sac of a pawn to open a file and come with a second rook at the 7th (2nd) rank. Maybe some programs "solve" just by calculating far ahead, and already see a forthcoming win of material too. But I judge by the results, not by the method how the programs achieve it. Regards, M.Scheidl
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.