Author: Dann Corbit
Date: 23:09:55 10/25/02
Go up one level in this thread
On October 26, 2002 at 00:42:16, Russell Reagan wrote: >In reading one of those old posts, Christophe Theron was talking about how he >did his pawn evaluation in a pawn hash table. I'm curious if this is a good >idea. There are obviously far fewer pawn formations than there are chess >positions. It seems unsound, but then again so did a transposition table using a >hash table implementation at first thought. Pawn hash is an old idea and almost all of the good programs do it. Score the pawn formation separately... Usually, the king is included in the pawn hash as well, since their relationship is very important. You don't score the pawn attacks at this stage... Just look for things like: 1. Passed pawns 2. Doubled pawns (and tripled pawns) 3. Pawn islands 4. Opening in pawns shelter for the king 5. Isolated pawns None of that stuff changes for all the things the other pieces on the board do, so you can hash it. Small subset of programs using Pawn hash: Amateur Amy Arasan AX Baron Beaches Bestia Betsy BremboCE Butcher Crafty Crux Delfi Esc Freyr Gargamella Lambchop Leila Madeleine Monsoon ExChess Gerbil GnuChess Gullydeckel KNIGHTD3 Pepito Resp Sjeng SSEChess TAO The Crazy Bishop Lots more besides these, I imagine -- I got tired of looking
This page took 0.32 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.