Author: Bruce Moreland
Date: 14:21:09 08/08/01
Go up one level in this thread
On August 08, 2001 at 15:31:53, Frank Phillips wrote: >Thanks for the full response Bruce. The commets on your test methodology are >valuable. > >I certainly was not picking on Ernst. Given the published results of his >systematic tests on a variety of test suites, R=2/3 looks attractive. My point >was really slightly off topic: about how exactly you evaluate the benefits of >changes. In my case I think with R=3 I see more tactical failures when playing >Crafty or the like, but doubt these are due to more aggressive null move. More >likely they can be reduced by removing unhelpful extensions, improving king >safety (in many cases) and speeding up the search. > >The other thought was that, logically, if the increase from (R=1 to R=2) was >good at some time in the past, then the increase from (R=2 to R=3) must also be >good at some stage. Presumably when speeds have increased the average depth >sufficiently to offset the side effects. It may be true that R=3 is good at some search depth. I had not considered that. bruce > >Frank
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.