Author: Eugenio Castillo
Date: 16:37:58 04/28/98
Go up one level in this thread
On April 28, 1998 at 16:45:03, Jon Dart wrote: >I do all of the things you mentioned, plus I have some >special code to evaluate multiple passers (sometimes the >opposing king could catch one of them, but not both), plus I >now have code like Crafty's to score "outside" passed pawns. > >The real problem is the magnitude of the scores. In the >endgame, it is fairly safe to give a big score for passed >pawns, especially if you have the knowledge items mentioned >above. > >In the middlegame, it is a lot trickier. Sometimes >an advanced passed pawn is just deadly, and eventually the >side that has it will pile enough pieces on it or around >it to force it through to queening, or force the opponent >to sac something to take it off. I get killed this way >pretty often > >Other times, the program pushes up a pawn to the 6th, and >eventually loses it. > >If you push up the passed pawn scoring too high, you will >find your program giving up material to force a passer, >and sometimes it will win that way, and sometimes it will >lose. > >I currently value passed pawns a little more highly as >material decreases, but the scores are not very big until >the endgame is reached. > >--Jon > >On April 27, 1998 at 18:41:28, Stuart Cracraft wrote: > >>What do people generally do to handle passwd pawn >>evaluation? >> >>A few things that could be considered: >> >> - square-of-the-pawn logic applicable for passed pawn? >> - squares in front of the passed pawn (especially queening >> square) controlled by whom? >> - passed pawn's rank >> - passed pawn's progress blocked by an enemy piece on same file. >> - phase of game's effect on value of passed pawn >> >>But what about prioritizing and valuing all of these? I'm looking >>mostly for guidelines and rules-of-thumb as I haven't seen too much >>in the literature about passed pawns handling >> >>--Stuart I think passedpawns evaluations must be different depending of the endgame type. It's not the same to be in_square (GNU concept) in an KP endgame than in a KRP one. Weak pawns could be more "weak" in KP etc.... In my program I use some special endgame rutines for KP, KRP and another one for more pieces (I think this three groups are the most important). Often ocurrs that after a piece change you fall in a simple lose KP position (and one move before it was a draw), the difference between a program who identify this and one who don't do it is great in endgames. Eugenio.
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.