Author: Uri Blass
Date: 11:56:39 02/20/05
Go up one level in this thread
On February 20, 2005 at 14:31:44, Jonas Cohonas wrote: >With all the parameters of Chessmaster (TheKing) i think it has been shown at a >pretty convincing level, that with few or many tweaks to the respective >parameters can improve the play of the engine, some settings work better for >blitz, some better for long games, some better for overall play and some better >for playing against a specific engine. > >With this knowledge, assuming it is somewhat accurate, would it not be possible >to make an engine that would be better for long analysis/games? > >I have asked this question before, but that was before the success of all the >different CM tweaks. > >Now one way to do this i guess would be to look at all the different aspects >that is needed for an engine to do well in short time controls and then replace >them with code that might have been left out for the benefit of overall strength >at most time controls? > >What approach would a programmer take to chess programming, if the actual kN/s >was not a big issue, say if an engine would normally do 800 kN/s and a >programmer added so much knowledge it dropped to 30 kN/s, that would probably >mean that it was no blitz monster, but if you gave it time it might come up with >better moves with the added knowledge? I do not think that generally adding knowledge to the evaluation is good for long time control and bad for blitz. The problem of chess programmers is to decide if something is productive and it is possible that knowledge is simply counter productive even if you assume the same number of nodes per second. Uri
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.