Author: Robert Hyatt
Date: 12:17:03 07/23/98
Go up one level in this thread
On July 23, 1998 at 13:48:10, Bruce Moreland wrote: > >On July 23, 1998 at 12:58:20, Robert Hyatt wrote: > >>The major move that Crafty didn't like was the Qh5+ move which traded the knight >>on h8 for black's remaining two pawns. Crafty has specific eval code that says >>if it is down a piece, even with three pawns for it, it isn't going to be happy >>unless all the pieces are gone except for that one extra piece for the opponent. > >Black got two pieces for a rook and three pawns. You consider this being down a >piece? > >bruce I don't count the pawns in this piece of code. I look and see "rook" for "two pieces" and say "*bad*" for the rook side. Ditto if I get a piece for two or three pawns. I say "bad" for the side with the pawns. How bad? Varies, but the penalty is over a pawn at present, to avoid doing the ugly B+N for a R+P, commonly on f2/f7, or a knight for 2 or 3 pawns to create some passers as well. This penalty scales down (piece for pawns) as pieces are traded, because I'd much rather have 3 pawns against a knight or bishop than have the bishop and only be able to hope for a draw. Ed faked around with the vaue of a pawn to avoid the piece for pawns trade. I used to muck around with the piece and pawn values to discourage this. But I got tired of the problems that caused, and, in fact, the current "penalty" is exactly how I do this as a human. I (and Crafty) would gladly sac a piece for two or three pawns around the king if there are attacking chances. But not just to call it "even" since the extra piece becomes a problem... So I consider N for 3P's to be +1.00 for the side with the N, if there are no other positional factors to change this. I consider a R vs BN (or any two pieces) to be -2 for the R side (-1 for normal material difference, -1 for what I call a "bad trade")...
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.