Author: Peter McKenzie
Date: 14:37:27 09/16/98
Go up one level in this thread
On September 16, 1998 at 17:16:20, John Coffey wrote: >On September 15, 1998 at 21:11:54, Robert Hyatt wrote: > >> >>1. you really can't extend forever, because the most captures you can >>ever have in a row is 30, because then you are fresh out of piecs to >>capture. >> >>2. I don't "extend" on captures as I think you are suggesting, although I >>do follow captures in the "quiescience search" but there captures are *all* >>I consider which limits this quite effectively... > >Thanks for the reply. Let me see if I understand this correctly .... > >You search out to a certain depth, and then beyond that you look at >every capture until there are no more captures to be made? > >A "quiescience seach" is an attempt to determine if the position is still >active? Thats right - pretty much 'everyone' uses a quiescience search which is an attempt to search until the position becomes 'quiet'. The idea is that once you reach a certain depth of full width search, you go into the quiescience search. In the quiescience search, the side to move has the option of accepting the current evaluation (standing pat) or playing a capture move. So if the static evaluation is above alpha, then set alpha to be the static evaluation. If the static evaluation is above beta, you have a cutoff. There are many slight variations of the quiescence search, for example: - some don't allow standing pat when in check - some try to include checking moves as well as captures - some try to only look at capturing moves that appear to win material - some futility prune capturing moves (no point capturing a pawn when you are a queen down as the other side will just stand pat next ply) - most include promotions (to a queen only usually) - some include pawn pushes to the 7th Which ever variation you use, good move ordering will make quite a difference to its performance. Cheers, Peter McKenzie
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.