Author: Robert Hyatt
Date: 08:29:18 06/06/05
Go up one level in this thread
On June 06, 2005 at 07:13:23, Steve Maughan wrote: >Andrew, > >One technique which could adds some value is to never do a hash cut-off at a PV >node (i.e. when alpha != beta - 1); and start to pull the evaluation closer to a >draw score when evaluating i.e. taking the 50 move rule into account. Since >there will not be a hash cutoff at a PV node, the PV and it's score will not be >polluted by the invalid hash scores. One effect may be a somewhat unstable >search i.e. the PV score may start to approach zero and you may get quite a few >fail highs for other moves that fail low on a research but this is probably a >price worth paying. > >Steve Doesn't help much, as almost all hash hits are not "exact" but are "boundaries". You get wrong scores on everything but the PV, which means that the PV itself is also wrong since it is relative to the non-PV scores that would be wrong.
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.