Author: Tim Mirabile
Date: 14:40:11 11/30/99
Go up one level in this thread
When trying to defend this kind of ending, it's important to use the proper "man-to-man" coverage. :) In this case, the black king can defend against a wrong rook pawn, but only if it stays within range. So it's important to assign the king to stop the rook pawn and the bishop, which could sacrifice itself for either pawn, to stop the other pawn. Can a program be "taught" this? Maybe in this specific case where the wrong rook pawn is involved. But I don't understand why the programs cannot see that once the black king gets to h8 it's just a simple draw. Some kind of horizon effect I guess - a human can easily see that no matter how long the white king runs around the best it can do is give up the d-pawn to win the black bishop, resulting in a draw. It is this kind of logic that the current programs don't have - some way to tell that something can never happen no matter how many random moves are thrown in. I am also surprised at how many trust the program's misevaluation of the position so much as to doubt their own judgement. On November 30, 1999 at 15:43:23, Howard Exner wrote: >I wonder if programs can be taught this theme. Most already understand the >rook pawn draw with the bishop of the wrong color. Can they code in something >to cover the above example? - if the defending bishop covers the queening square >of the other pawn(the d7 one in this case) then eval it as a draw(that is when >the defending king reaches the >drawing zone for the rook pawn). Humans spot this easily but can it be >programmed easily?
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.