Author: Robert Hyatt
Date: 12:58:02 03/22/00
Go up one level in this thread
On March 22, 2000 at 13:50:51, Alessandro Scotti wrote: >On March 22, 2000 at 10:21:07, Inmann Werner wrote: > >>Hello. >> >>I once tried the fractional extensions, and got really bad results, although in >>Crafty they work perfect. >>If I extend on check lets say 0,75, there will be no extensions since there are >>two checks, and that may be out of horizon. >>If I extend 1,5 at check, i get wonderful results at some testsuites, but a >>really poor performance overall. >>How do you manage, that the fractional extensions work, when they are needed, >>without exploding the tree? A extension lower then 1 is no extension at this >>ply, and if it is the last one before quiet search, you are lost?!?!? > >I think that extensions less than 1 are useful when you want to take note of a >fact that, considered alone, isn't enough to trigger a full ply extension but >might if the fact itself repeats later in the search or several of these facts >occur at the same time. >If the sum of such extensions is less than 1 and you are just starting to >quiesce then there is no extension and that seems correct, meaning the fact or >facts are not interesting enough at this point. That is the main point, although the extensions carry across plies as well. IE 3/4 at depth=N, plus 1/2 at depth=N+x will extend at depth N+x since that rolls over the 1.0 extension threshold...
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.