Author: José de Jesús García Ruvalcaba
Date: 12:08:07 04/14/99
Go up one level in this thread
On April 14, 1999 at 13:09:44, Bruce Moreland wrote: > >On April 14, 1999 at 12:15:35, KarinsDad wrote: > >>I couldn't have said it much better. The software I am working on has many of >>the elements that you describe. How well it will work is yet to be determined. >> >>However, one other vast area of improvement that I foresee is the endgame. >>Having tablebases is fine, however, a tablebase is no different than a typical >>player's slight endgame knowledge (the tablebase is just normally more >>complete). Once programs really understand endgame concepts, I think that their >>level of play will improve dramatically. > >Endgame databases help in specific very low material situations. These >situations come up, but there is usually a lot of endgame play before they come >up. > >Sometimes they can help in situations with slightly more material. An example >is KRPP vs KR, where the pawns are on the a- and c-files. This ending can be >hard to handle, but the KRP vs KR table lets the program understand the >consequences of exchanges, which are critical in this case. I've had either >side of this come up, and I've been satisfied with the result. > >There are plenty of endings where endgame databases have no bearing on the >outcome. > >People often talk about the ending as a place to add sophisticated knowledge, >but I've never heard an example. > >bruce In Computer Chess II, by David Welsh and Boris Baczynski (hope to spell it correctly), a program named PAWNKING (by Helmut Horacek) is mentioned. It played only pawn endings, and it seems that it had sophisticated knowledge. I would propose different evaluation functions for different type of endings. One for pawn endings, one for same-coloured bishop endings, one for opposite-coloured bishop endings, one for knight endings, one for rook endings, one for bishop versus knight endings, and one for queen endings; and some more for endings with more material. The evaluation function can be changed according to the pawn distribution (all on the same side, pawns on both sides, or with pawns scattered all around the board), and the presence/absence of passed pawns.
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.