Author: Alvaro Jose Povoa Cardoso
Date: 01:27:51 08/08/01
Go up one level in this thread
On August 07, 2001 at 23:08:14, Robert Hyatt wrote: >On August 07, 2001 at 16:42:27, Alvaro Jose Povoa Cardoso wrote: > >>Do you still have the ETC version of Crafty? > > >No. But here is what you need: > >1. A function that will compute a updated hash key given a specific move. >This is needed to avoid the MakeMove() overhead that will arise when running >thru the move list to find if any of the resulting positions are hash hits. > >2. A simple ordering function that takes the move list, makes them one at a >time, and probes the hash table to see if you get a hit and that hit will be >an exact score that causes a cutoff, or if it will directly cause a cutoff. >If so, that is the move you try first at this ply. > I thought I could return immediately! >>If not could you post the ETC code/pseudocode, if that is not asking too much? >>I know the net result of implementing ETC is null, but I think my program >>(checkers) could benefit more than chess programs from ETC (at least I would >>like to test). >>I'm confused with ETC. Taking Crafty as the example, should we use HashProbe() >>with ETC or should we make a special version of HashProbe() that does not >>changes alpha and beta since all we want is the hash score? >>Correct me if I'm wrong: >>When using ETC, after making each move we should probe the hash table for the >>opposite side right? > > >Yes, as you are effectively "making" a move. > > > >>So instead of >> HashProbe(tree,ply,depth,wtm,&alpha,&beta,&mate_threat) >>we should do >> HashProbe(tree,ply,depth,ChangeSide(wtm),&alpha,&beta,&mate_threat) >> >>Cutoffs should be done if the hash score >= BETA right? >>When probing the hash table for ETCs should we consider >>Upperbounds/Lowerbounds/Exactscores or >>sould we simply use the hash score? >> >> >>Thank you, >>Alvaro Cardoso
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.