Author: Robert Hyatt
Date: 17:16:10 08/18/98
Go up one level in this thread
On August 18, 1998 at 16:48:58, vincent dichiacchio wrote: > >could someone explain some of the methods that limit search extensions from >exploding. i have heard about fractional ply, but that doesnt seem much >different than setting an absolute limit on extension. is the difference more >evident when extensions extend on top of extensions? are the methods arbitrary >just because its too costly to implement something more selective? > >thanks in advance, >vince The basic plan is to effectively limit extensions to no more than one ply of extensions for each ply of search. That might produce a non- terminating search for checking lines, but hopefully those get clipped by repetition detection. But by limiting the max extension at any ply to one ply, and knowing that it is very unusual to extend every ply (except for a series of checks where the opponent has only one legal way out of check which also extends) this will terminate. And with fractional ply extensions, I control this better by using "check=1 ply" but one-legal- response to check can only extend 3/4 of a ply. Which means that 3 of every 4 one-response-to-check moves extend, but the 4th one doesn't... which also limits the depth nicely... this has been used for as long as I can remember.. Belle used it in 1983, for example. Cray Blitz was using it in 1980 at least, maybe earlier. And it wasn't *my* idea either... possibly Dave Slate, Ken Thompson or Murray Campbell, all of whom I talked to regularly back then...
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.