Author: Ernst A. Heinz
Date: 16:23:41 03/02/01
Go up one level in this thread
>>Easy -- the relative cost of a depth-reduced null-move search >>at frontier nodes (with a remaining depth of 1 ply) is just >>too high. Most normal searches from there also go straight into >>quiescence mode and the gain of null-move cutoffs does not seem >>to outweigh the added null-move search effort. > >Yes, this seems to be the case - has this been tested in even higher remaining >depths (2,3...)?. Maybe in some situations your could turn null moves off at >pre-frontier nodes also? After all, this would lead straight to qsearch? Maybe >limiting null moves to levels where remaining depth after the depth reduction is >at least 1 to allow any kind of move for opponent. I'll have to test this. > >I tested the version that disables when d==1 and got 0-10% reduction in the node >count (only in a couple of positions...) so this seems to be a very good idea. Yes, almost everybody seems to disable null moves at frontier nodes -- including "DarkThought"! Disabling null moves at pre-frontier nodes does not look so clear-cut because there most normal searches do _not_ hit the quiescence mode directly. Hence, the null-move overhead still seems to pay off at pre-frontier nodes. =Ernst=
This page took 0.03 seconds to execute
Last modified: Thu, 07 Jul 11 08:48:38 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.