Author: Andrew Williams
Date: 02:18:34 07/26/02
Go up one level in this thread
On July 25, 2002 at 18:26:40, Scott Gasch wrote: >Hi, > >I recently reworked my hash tables to be more efficient and found, to my >surprise, that my engine could not longer solve FINE70. Well I hunted this down >and found that it was not a hash bug but rather a problem with an addition I >made a while ago: a nullmove verification search. The idea (which may be a bad >one, it seems) is to _always_ nullmove (even in late endgames) but before >accepting a FH after null search, do a reduced depth search to verify the FH >result. > >It seems like this works very well in some positions (and has the side effect of >allowing you to do a zugzwang extensions when the nullmove search FH and the >verification search does not). But it royally screws up the FINE70 position. > >My question for other programmers around is: is there anyone else playing with >this null verification search and how are your results / impressions? > >Scott At the moment, standard PM gets Kb1 at depth 25 (~4M hash entries). But normally I don't allow null move when there are 0 pieces for the side on move. If I *do* allow this, PM needs until depth 28 to find Kb1. Andrew
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.