Author: Dieter Buerssner
Date: 16:32:32 10/13/03
Go up one level in this thread
On October 13, 2003 at 18:09:36, Steve Maughan wrote: >Another reason why it might not be sound is if a program has large speculative >positional terms. For example if a program give a large penalty for a trapped >rook (e.g. penalty = 4.5 pawns) then it is likely to be much stronger at slower >paced games since the added depth will mean that the terms only kicks in when >the rook is truely trapped. In contrast if the term is used at fast time >controls the evaluation may conclude that a rook is trapped when there is >actually a way out - leading to such plays as knight sacrifices to 'save' the >rook. I agree. I think, however, that the same problem can show itself in the different direction - the engine with more "knowledge" may get weaker with longer TC. Assume some positional terms, that the search would see few plies deeper. Now, the knowledge might not be perfect. The knowledge (paired with pruning decisions in the search) might hide the exception to the "clever" engine, while the engine without the knowledge might not have any problems to see far enough with deep search in the other cases, where the knowledge would have been correct. For me, it is far from clear, that more knowledge should mean better at longer TC. Regards, Dieter
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.