Author: Gian-Carlo Pascutto
Date: 08:57:53 01/05/01
Go up one level in this thread
On January 05, 2001 at 11:22:14, Dieter Buerssner wrote: >I think, this can still happen, when you use pruning techniques, that >depend on the window. I did not use anything besides nullmove in my test. >On such technique is null move. Perhaps with >beta 1.0 the null move will fail high. With 1.1 it won't fail high anymore. >You would no assume, that later in the "normal" search, you get a score >= 1.0. >But this might not be the case, because actually the null move refutation >would have been wrong for 1.0 allready (perhaps because of some slight >Zugzwang). I can reproduce it quite easily in WAC 1 for example. If I set the bounds to [-inf,inf] (no aspiration search) I will see the mate at ply 3 with the correct score. If I do a normal aspiration search, it will pick Ne8 at ply 3 and only see the mate at ply 5. At ply 3, it will fail high, then low on Qg6, and _not_ give a mate score. Do you mean this would be caused by the nullmove's score being used when searching with the narrow window, but not when using the full window? I.e. when you use the nullmove score you are relying on a (possibly incorrect) shallow search, whereas without the nullmove it would use a deeper search. This would seem to make sense (5 - 3 = 2 =nullmove-R-factor in this position), but then I do not understand why researching the position with the full window will not find the mate (at ply 3). >Also, when extensions depend on the window, such artifacts can arise. I did not have any of those in the tests. -- GCP
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.