Author: Bas Hamstra
Date: 13:59:50 12/24/02
Go up one level in this thread
On December 24, 2002 at 15:17:02, Russell Reagan wrote: >On December 24, 2002 at 14:00:22, Bas Hamstra wrote: > >>- - - - - - - - >>- - - B - - - - >>- - - W - - - - >>- - - - - - - - >>- - - - - - - - >>- - - - - - - - >>- - - - W - - - >>- - - - - - - - > >Saying that the white pawn (either) is weak is like two men having a duel, and >one man shoots the other, and the man that got shot feeling that it was a draw >because both men are going to die, even though his opponent may not die for many >years. > >The white pawn may, by definition, be "backward" or "theoretically weak", but, >all other things being equal, black will lose its pawn long before black could >exploit the white weak pawn. It's all relative really. If all other things are >equal, then white has an advantage. If black can capture the white pawn on d6 on >the next move, then it's most likely a draw. > >Instead of thinking, "is this pawn weak?", think, "is this pawn weak relative to >my opponents'?" The one pawn may be weak, but in comparisson to the opposition's >pawn structure, it may be a great advantage. This is just more of the danger (or >lack of perfection) in detecting this stuff statically. The evaluation might >say, "both sides have a weak pawn, even score" when that is not the case. >Black's pawn is weaker than white's. It is planted there forever, further >restricting black's pieces. Barring any immediate captures of white's advanced >pawn, white will eventually be able to trade off pawns and obtain a passer. >Black may be able to draw, depending on the other circumstances. > >The point is, again, that you will probably always be able to find specific >positions that your current method doesn't handle correctly. Look for the >general solution, then try to generalize it more. Detecting dynamics statically >is a difficult thing to do :) Maybe, but it works if you do it right. Take for example Crafty's candidate passer code. It provides a very effective look-ahead that causes it to put it's pawns right far before a search could ever detect it. Currently I am experimenting with detecting the set of pawns that is not pawn-defendable. At first sight these points in the position seem really interesting. And I don't have to penalize them all alike, it's easy to classify them further, for instance penalize "open" pawns that can't move the most, other open pawns in between en non-open pawns less. I fail to see why an undefendable open pawn is ANY better than an isolated open pawn. In short: worth a try for me... Best regards, Bas.
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.