Author: Roberto Waldteufel
Date: 15:52:08 10/03/98
Go up one level in this thread
On October 03, 1998 at 17:03:19, Inmann Werner wrote: >I read the article of Ulli Türke about razoring. I must say, I did not >understand it full, but I tried to implement something like this > > >In alpha beta after checking for checkmate and the extensions i try > > if (noextension_made && distance to horizon<=2 && material_bilance[depth] > + 6*pawn_Value <=alpha) return(alpha-1) > >In test positions,i tried, this decreases the tree about 15% with no bad effect. >But I am suspicious. > >I think the idea is. >I have done a move and my material_balance is much worse than alpha after doing >it, and now comes the opponent which will make it more worse....So forget this >move and do not go deeper. > >Is above buggy or too risky? > >Some suggestions? > >Werner This sounds very similer to Ernst's extended futility pruning in DarkThought. I don't see any problem, but I think Ernst uses a 2-pawn offset at depth 1 and an offset between the value of a rook and the value of rook+pawn. Your value of 6 pawns looks very similar to Ernst's value at depth two, but perhaps you could speed up even more by relaxing the threshhold for depth one. Best wishes, Roberto
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.