Author: Robert Hyatt
Date: 18:54:09 02/01/03
Go up one level in this thread
On February 01, 2003 at 17:38:00, Sergei S. Markoff wrote: >Hello All! > >False hash probes in endgames is well known problem of the classical search >scheme. Same position can be the result of move chains which has a different >lengths. It causes the misevaluation because of the 50 moves rule. > >The one idea is to use some multiplier N when evaluating position in which >50-moves rule counter is large. Fur example when this counter shows 50 plies we >can multiply evaluation onto 0.6 (or other value <1). Also we'll try to save >50-moves rule counter in hash entry and use this value to recalculate hash >evaluation. > You can fudge around all you want, but there is no way to solve the problem short of storing the complete path played to reach a position in the table entry, Which would make the table useless and way too small... You just have to ignore it. >I'm not implement this scheme at present but I'll try it in future. Are you >tried it? Or you have other non-secret approaches? I see that Ruffian is safely >avoiding this problem (see exact Ruffian evaluation in some positions, which is, >I think, not based on knowledge, for example: black: kh1, ph2, ka7; white: kf1, >wtm) (Fritz7 also shows 0.00, but immediately, I think it's because of specific >knowledge). > >Sincerelly yours, >Sergei
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.