Author: Robert Hyatt
Date: 13:37:36 12/18/02
Go up one level in this thread
On December 18, 2002 at 16:28:09, Dezhi Zhao wrote: >I read some days ago you experimented the verified null move idea with Crafty >and some encouraging results. However, have you decided to keep the change? > >I think the VR=3 search is more akin to R=3 search in the long run. This idea >seems to me is that you verify with reduced depth a likely cut node that was >detected by a fast R=3 null move search. Though cheap screening (R=3) and >expensive verification (reduced depth search) sounds a reasonable trade-off, all >the sub-nodes of the verfication are still searched by using the less-reliable >R=3 again. The verification could fail the same way as standard R=3 to some >less extent. I think the validation process of the verified null move >in a real program like Crafty could be lengthy. I haven't decided anything yet. It seemed better in some cases, worse in others, and overall seemed to be a "break-even" on the first implementation. I got side- tracked with the hyper-threading stuff and this new machine, and final exams and projects, so I haven't played with it any more recently...
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.