Author: Robert Hyatt
Date: 08:38:42 06/03/98
Go up one level in this thread
On June 03, 1998 at 09:52:34, Roland Pfister wrote: > >On June 03, 1998 at 08:31:28, Robert Hyatt wrote: > >>On June 03, 1998 at 02:45:34, Roland Pfister wrote: >> >>> >>>So do I. I just don't allow extensions if a nillmove is in the path. >> >> >> >>I do normal extensions and everything else, as I want that null move >>search to fail high... and the best way to do so is to follow checks and >>so forth to prove that "passing" is really ok, showing that the current >>position is good enough that even passing doesn't hurt. > >Sounds plausible. My idea is that I search nullmoves with reduced >depth and I don't want to undo that by following extensions. My tree >is smaller and I hope to get to the next iteration by doing that, so >that it will >compensate for reducing my depth because I did not follow the >extensions. > >Some time ago I did a statistic of cutoffs ( its actually in the >development >version of Patzer ) and noticed that nullmoves get the highest success >rate ( 70 to 90 % iirc ), so I never thought of changing my way to do >it. >When I find the time I will try both ways on BS2830 or LCT2. > >I think that I have read something like that ( we reduced the depth, >so we won't increase it by doing a extension ) in the Crafty source. >Could be a year ago or so. I'll look at the comments, but the search definitely doesn't know that there is a null-move in the history, except that it does know that the move at the previous ply was a null or not to avoid back-to-back nulls.
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.