Author: Dave Gomboc
Date: 15:03:49 04/15/99
Go up one level in this thread
On April 15, 1999 at 15:12:19, Bruce Moreland wrote: > >On April 15, 1999 at 11:29:00, Stuart Cracraft wrote: > >>On April 14, 1999 at 13:09:44, Bruce Moreland wrote: > >>>People often talk about the ending as a place to add sophisticated knowledge, >>>but I've never heard an example. >>> >>>bruce >> >>M Chess > >Right, and what is the knowledge? > >bruce Maybe people are thinking of stuff like "Dvoretsky's rule of five." (Endgame: having this distance between rook and the opposing king is important because then the king cannot force the rook away from pressuring something (e.g. a passed pawn) in the position.) Search would figure this out on its own if it goes deep enough, but the eval term will make software see whether the rook can hold its position earlier, which is always useful. There are any number of these "rules of thumb", that when applied at leaf nodes might improve a program. As you know, this is a long way from making a program "knowledge-based", which is a total buzzword when it comes to most of the software out there. Something like PARADISE is what I would consider knowledge-based, because in it, search is not actually in the driver's seat, something that probably can't be said for MChess, or Hiarcs, or Rebel, or CSTal either, for that matter. [Disclaimer: only the authors _really_ know..] Dave
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.