Author: Tord Romstad
Date: 10:11:17 01/31/04
Go up one level in this thread
On January 30, 2004 at 16:09:51, Ed Schröder wrote: >I can imagine my way of doing things is confusing, maybe when I tell a little >about Rebel's history you will understand better. > >In the past (before 2001 or so) Rebel was a full static selective search >program. Every node in the tree is evaluated and the score is used as a base to >prune the tree. > >And now nullmove moves in, in the case the selective search returns the message >to prune I call nullmove with as only reason to verify if it is safe to prune. >If nullmove returns a score >= alpha the node is searched after all. So I use >nullmove to verify my static selective search. Thanks. Seems rather similar to the "lazy verification searches" I experimented with recently, which I ended up discarding. I should probably give it another try. >Nullmove is only practiced in the very few plies of the search, iteration >driven. It takes out the worst failures of my static selective search. > >With a simple parameter I can make Rebel a standard R=2 or R=3 program, it then >runs 2 times slower and loses considerable against the default setting. > >Hope this helps. Yes, I hope I understand it better now. I hope to find the time to do some experiments next week, and will contact you again if I have further questions or discover some interesting ideas Thanks for your help, Tord
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.