Author: Robert Hyatt
Date: 10:35:48 07/09/02
Go up one level in this thread
On July 09, 2002 at 03:32:34, Christophe Theron wrote: >On July 08, 2002 at 16:24:32, Steve Maughan wrote: > >>Christophe, >> >>[SNIP] >> >>>It doesn't matter to what depth you are searching. >>> >>>Increasing R can in the best case only give a minor speedup. >>> >>>The speedup you can get by increasing R is bounded to a few percent. >>> >>>Just evaluate the time you are spending in null move searches currently. >> >>[SNIP] >> >>Is this really true. I would say that to measure this you'd need to see what >>percentage of leaf nodes have one or more null moves in their path. I'd be >>surprised if it were only a few percent. >> >>Regards, >> >>Steve > > > >It's the whole point of null move. > > > > Christophe Actually it isn't. The point of null move is to replace a larger search with a significantly smaller search. IE searching a real move with depth D, vs a null-move with depth D-R. The number of null-move searches is _very_ large since it is done first at almost every ply in the tree. And the number of nodes below null-moves is typically 1/3-3/4 of the entire tree space searched. Reducing that space further will be a huge savings.
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.