Author: Thorsten Czub
Date: 23:48:21 05/07/98
Go up one level in this thread
On May 07, 1998 at 19:41:54, Don Dailey wrote: >I don't see how this can be easily progarmmed without knowing genius's >evaluation. Cstals evaluation is unreliable in the sense that it >doesn't >measure the positions goodness conservatively, in the same way other >programs do (at least relative to cstal.) If the programmer used a >conservative version to test for "winningness" it's defeating the whole >idea not to mention really expensive. The whole problem is judging >when to switch over and I'm guessing Genius evaluation is the best >measure of this! But don't ask Cstal when to switch over. Exactly ! In Paris and also at home I do always breathe out when CSTal attacks in the moment THE OPPONENT goes into minus evaluation. In paris they (Dieter or Frederic) made jokes about me always starring like a hypnotized/mesmerized rabbit on the screen of the opponent. THERE the important data was to find. >I'll bet this wouldn't improve Cstal's results though. I think it's >much more productive to have the programmer add more knowledge about >when to be conservative and when not to. > >By the way, I too am impressed with the way Cstal plays chess. It >seems to know when to play an unsound attack and get away with it! >At the Aegon Cilkchess played a few games with it and didn't do very >well. In most of the games Cilkchess was winning until Cstal made >an UNSOUND winning move! However, I want to mention that CSTAL had >a big hardware advantage, I was running on a slow laptop with the >serial version of Cilkchess, tiny hash tables and no thinking on the >opponents time. The hardware was about 3 to 1 in favor of CSTAL. >But I could see that CSTAL was not weak and played extremely provocative >chess style. Much nicer than Cilkchess's boring style. Maybe one day we meet us (e.g. in netherlands - when I find a dutch aunt or learn dutch) the netherlands championship and we can try out THE BIG MACHINES. I am sure Cilkchess kills us than ! CSTAL can bluff, but not against a much stronger opponent. Chris and I work on the Windows version. In it we can better implement the 2-engine concept or the CONSERVATIVE score to handle this problem. We will see where we end... > >- Don
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.