Author: Gian-Carlo Pascutto
Date: 02:46:47 10/12/01
Go up one level in this thread
On October 11, 2001 at 16:52:49, José Carlos wrote: > At a certain position in qsearch, you know the material balance (most of us >calculate it incremental), but you have no idea about positional score until you >call Evaluate(). If material is very low compared to alpha (for example, >alpha==0 and material==a_queen_down) then you know the positional score won't >rise the evaluation of the position above alpha, so you don't bother calling >Evaluate(), just search the captures (this is up to your implementation of >qsearch). > Also, if material is far above beta, you can fail high on that node, because >you know the positional score won't get the eval down enough. This description is what I would consider futility pruning, not lazy eval. Lazy eval works by sorting out your eval from fast/big score to slow/small score and adding in a few checks if the remaining terms can bring the score within the alpha/beta window. -- GCP
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.