Author: Aaron Tay
Date: 03:32:12 01/07/03
Go up one level in this thread
On January 06, 2003 at 18:00:04, Terry McCracken wrote: >On January 06, 2003 at 10:16:22, Robert Hyatt wrote: > >>On January 06, 2003 at 09:36:53, Bob Durrett wrote: >> >>> >>>The concept of "trading down to a won endgame" and similar ideas such as >>>avoiding a dangerous attack by trading down to an equal or slightly inferior >>>endgame seem conceptually simple. >>> >>>Endgame knowledge is available everywhere. For example: >>> >>>The recent publication, "Fundamental Chess Endings" by Karsten Muller and Frank >>>Lamprecht, copyright 2001, Gambit Publications, Inc., ISBN 1901983536, includes >>>a table at the back of the book titled "Table of Computer Database Results for >>>Pawnless Endings." This table provides "endgame knowledge" information, which >>>could be [and probably is] used by chess engine programmers. Similar >>>information appears elsewhere in the written chess literature. >>> >>>A chess engine position evaluation could check to see if the position matches up >>>with one of the known endgame types. [This might take special dedicated >>>software.] Then the knowledge could be used to determine whether or not to >>>exchange down to that endgame. I'm sure [??] that modern chess programs do >>>something like that, although perhaps not aimed at making an "exchange down or >>>not to exchange down" type of "decision." >>> >>>The idea that a position would receive a numerical value is common. But is it >>>really necessary that search choices be based solely on such numerical >>>evaluations? Perhaps modern chess engines are not that simplistic? >>> >>>In summary: I don't see why a chess engine should have any trouble at all in >>>positions where exchanging down to an endgame is indicated. What am I missing? >>> >>>Bob D. >> >> >>Nothing. This is done by many programs now, mine included. Not all understand >>key ingredients such as distant pawn majorities, but most recognize classes of >>endgames that are favorable to them... > > >Hi Bob, > >Check my post. > >I think it's important for machines to be given lessons, beyond EGTBS for >such positions. > >The book is wrote, we don't neglect it, and programmers work hard on the >middlegame, but these relatively simple K+P endings get little attention. > >I haven't tried Crafty on this following position. > >http://www.talkchess.com/forums/1/message.html?275232 > >Terry More damaging is that crafty fails the second one here http://www.talkchess.com/forums/1/message.html?275366 Very disappointing given that crafty is reputed to understand the value of distant passed pawns and potential 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.