Author: blass uri
Date: 00:06:44 07/31/99
Go up one level in this thread
On July 30, 1999 at 17:21:48, José de Jesús García Ruvalcaba wrote: >On July 30, 1999 at 16:08:44, blass uri wrote: > >> >>On July 30, 1999 at 14:05:01, José de Jesús García Ruvalcaba wrote: >> >>> A few weeks ago somebody asked which selectivity levels were better for CM6000 >>>to analyze at very long time controls, suited for correspondence play. >>> The advice given was to use zero selectivity. I do not remember why I did not >>>answer at that time, but my advice is exactly the opposite: the longer the time >>>control the higher the selectiviyy level should be set, and not only for CM6000 >>>but for any program that allows you to change it. >>> The reason is that, due to the exponential nature of the search tree, the extra >>>time is almost worthless when using no selectivity or very low selectivity. >>>Jos >> >>It depends on the program. >> > > Yes. I was trying to state the general tendency that higher selectivity is >better for longer time controls. > >>I know that previous version of Rebel were more selective and it cause a problem >>because Rebel prune some good moves with the selectivity. >> > > Any selectivity can (and will) prune some good moves. But the risk should be >minized at longer time controls. > >>If you use high selectivity (do not analyze moves that you consider as illogical >>moves) then you can miss simple tactics because you prune some good sacrifices >>and do not analyze them and it may be a big problem at long time controls. >> > > I think that kind of selectivity (not analyzing "illogical" moves) is not used >since the late seventies. I meant selectivity that analyzes moves with poor >scores less deeply than moves with good scores. Analyzing moves with poor scores less deeply will cause your program to be a slower solver and chessmaster with ss=10 is not a slower solver so it probably does not analyze moves with poor scores less deeply. 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.