Author: Anthony Cozzie
Date: 05:27:17 11/22/03
Go up one level in this thread
On November 22, 2003 at 08:22:47, George Tsavdaris wrote: >On November 22, 2003 at 05:09:45, Gian-Carlo Pascutto wrote: > >>On November 22, 2003 at 05:05:59, Jorge Pichard wrote: >> >>>3.4 million nodes per second in the starting position, with an 18-ply search >>>depth in less than four minutes! Experts will recognise that this is quite >>>extraordinary. Frans Morsch and Mathias Feist, the authors of the program, tell >>>us that Fritz has never before run on such a fast machine. >> >>Don't forget thats 17 ply with hard forward pruning and slim eval >>vs 17 ply without forward pruning and a huge eval. > >I know very little about forward pruning/evaluation ..., so please tell >me: When you say Diep is searching 17 ply, do you mean it examines all the >possible positions for the following 17 plies, even the stupid moves? >(With a small calculation i think this is impossible but...) >Or that Diep discards a lower percentage of not-promising moves? > >Good luck to you and to all of the participants. > > >>So far I'm impressed with Diep, I didn't really believe Vincent >>would get things working so well. We'll see how well it really works >>in the games of course. >> >>-- >>GCP Well don't forget alphabeta. Diep also uses nullmove pruning (same as Fritz). What GCP is referring to is the program making some direct calculation (no search) and deciding not to search a child node. Example: I am down a queen, and this move hangs a rook, and the depth is low. I'll guess this move is bad. Now, most of the time this is good, but in WAC141 . . . . Anyway, there are countless methods for forward pruning, and they are all secret :) Shredder is doing mad forward pruning, and I think Sjeng forward prunes as well? anthony
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.