Author: Tord Romstad
Date: 11:18:32 06/17/04
Go up one level in this thread
On June 17, 2004 at 14:12:14, Fabien Letouzey wrote: >On June 17, 2004 at 13:49:46, Robert Allgeuer wrote: > >>What is the "blemish effect"? Is it a jump in the score? > >Yes, due to a different way of scoring features in position "classes" (e.g. >midgame/endgame in chess). > >>Why is it actually bad to have discontinuity in the scoring function? > >>Thanks >>Robert > >Because the program might think that it can stay in "class A" where the score is >good, whereas it might only be delaying an unavoidable transition to "class B" >(actually bad). It is a sort of horizon effect between classes. Or the similar case where there isn't really any big difference between class A and class B, but (because of an imperfectly tuned eval) the engine *thinks* class B is bad, and makes all sorts of serious positional concessions in order to avoid class B. Tord
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.