Author: Uri Blass
Date: 15:59:42 08/01/04
Go up one level in this thread
On August 01, 2004 at 18:17:41, Michael Henderson wrote: >I started a previous thread on this about my program getting 86% for this. >Everybody's comments were very helpful and I'm working on it, although I'm not >sure I'm even accounting for this stat correctly! Should I count null moves as >fh first(not doing that currently)? I'm counting all the other moves though >(except hash cutoff/move). The way the program is now, I get 97% with null move >off, 86% with it on. If I modify it to count null moves as fh first, i get 97%. > Comments appreciated :) > >Michael I do not understand why do you care about it. I do not see how knowing the statistics can help me to improve my program so I simply do not care about it. I think that good pruning can reduce the statistics without reducing the quality of order of moves because by not searching lines that you are sure that you are going to fail high you have relatively less fail highs. If you search after 1.e4 Nf6 2.Qh5 then you have a fail high in the first move that is 2...Nxh5 you may also have first move fail high for black later in the tree and you have no choice because after 1.e4 Nf6 2.Qh5 Nxh5 3.a3 every move of black is going to fail high. If you are smart enough to return beta after 1.e4 Nf6 Qh5 without searching then you have not first move first fail high in all these cases inspite of the fact that your order of moves has the same quality(in the case of null move pruning you are going to have fail high for 2...Nxh5 but you are not going to have fail higjh for 1.e4 Nf6 2.Qh5 Nxh5 3.a3 because 3.a3 is pruned by null move pruning. Uri
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.