Author: Bert van den Akker
Date: 14:47:51 11/19/00
Go up one level in this thread
On November 19, 2000 at 15:44:09, Bert van den Akker wrote: >On November 19, 2000 at 10:19:44, Uri Blass wrote: > >>On November 19, 2000 at 09:10:12, Bert van den Akker wrote: >> >><snipped> >>>I think a brute force program with only null move pruning added is at least >>>100 elo points stronger the pure brute force program because it will >>>search 1 or 2 ply deeper. >> >>There are lines that you are going to see deeper without null move because null >>move pruning prunes them so you cannot use the difference in plies in order to >>calculate the difference in rating. >> >>The only way to know is by playing games. >> >>Uri > >In this respect you are right. >But there are also lines which you will see deeper with null move pruning >because the general search will be deeper. > > >My experiments with null move pruning shows that null move pruning >can cut 90 to 95% af the search tree. In general some combinations >are seen 1 or 2 plies later with null move pruning but this can be compensated >with threat and other extensions and the fact that you are searching 2 plies >deeper in the same time. >A bigger problem with null move pruning is that moves whitch gives subtile >positional advanges can be pruned because of the nature of the null move >method. > >So making a very advanced positional evaluation makes not mutch sense if you >don't make some provisions. > >Maybe this is the reason why in the null move pruning is not used. > >BvdA The last sentence must be: Maybe this is the reason why in REBEL the null move pruning is not used. A small test with the WAC test positions gives the following results with my program: 5 seconds/move No null move pruning 300/281 10111111111111111111 11111111111111111111 11111111111111111111 11111011111111111111 11111111111011111111 01111111111111101111 11111111111111111111 10111101111111111111 11011111111111111111 11111111111111101111 11111111111111111111 01111111001111110111 01111111111011111111 10111111111111011111 11111111111111100111 With null move pruning 300/285 10111111111111111111 11111111111111111111 11111111111111111111 11111011111111111111 11111111111011111111 01111111111111111111 11111111111111111111 11111101111111111111 11011111111111111111 11111111111111101111 11111111111111111111 11111111001111110111 11011111111111111111 10111111111111011111 11111111111111100111 BvdA
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.