Author: Vincent Diepeveen
Date: 14:12:16 04/09/03
Go up one level in this thread
On April 08, 2003 at 12:02:09, Sune Fischer wrote: >On April 08, 2003 at 07:30:24, Bo Persson wrote: >> >>One of the things Aske Plaat discusses is that the "minimal tree" isn't the >>minimum! With hashing it is possible to get a search tree that is actually >>smaller than the alpha-beta theoretical limit. >> >>If you search with variable depth, a non-uniform branching factor, hashing, and >>null-moves, etc, reaching 1.21x the minimal tree MIGHT not be good enough, when >>it in fact COULD be possible to reach maybe 0.8x the "minimal tree". We don't >>know. That is a problem! > >I agree, but when talking about minimal tree in the classic sense I (and I >understand that this the common way) think of it as the best-case alpha-beta. >As I understood the question, hashing was also to be ignored. > >Hashing is an addition to the move ordering that can be "disabled" (just like >extensions and pruning), so you can measure move ordering relative to the >"minimal tree". >Hashing complicates matters "infinitely" here. :) hashing makes in diep the move ordering 1% better in fact (the positions which do not give a cutoff of course, they are disregarded in that measurement). In theory 1% better is not 1 ply deeper if you go do some math. But in fact 1% better move ordering means about 1 ply deeper search i found out by practical research. So clearly nullmove and transpositions play a major role. >-S. >>Bo Persson >>bop2@telia.com
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.