Author: Christophe Theron
Date: 19:34:26 11/26/97
Go up one level in this thread
On November 25, 1997 at 19:32:40, Bruce Moreland wrote: >I have no mobility terms in Ferret, it seems to do fine without them. >Sometimes it gets bad pieces, but so do other programs. > >I'm not opposed to mobility terms, maybe they work. If someone has >experience with them, or knows of a professional program that *for sure* >uses them (not just a deduction of the form, "they are very slow, so >they must use mobility"), please let us know. I would bet ChessMaster 4000 uses mobility. I've played many games against it, and it's completely obvious that it has huge weights for mobility. It has 2 main goals in the game: 1) crush your mobility 2) attack your king. Sometimes it can sac 1 pawn to get a large mobility advantage. I don't know about CM5000. Can someone tell us? >I think it may be possible to be more sophisticated than this, without >having to go to the trouble of iterating the pseudo-moves for the >pieces. The 32-bits version of Tiger that played the 15WMCC had no mobility evaluation, except in piece-square tables. And it is a bad way to do it. Take a look at the game Tiger-Stobor (11th round). Everything happened as if Tiger didn't play that game. I got 3 bad pieces! I've been so upset by this game (Tiger could have been in the middle rankings if only it had played a little bit), that my only work since the world championship has been to include a REAL mobility evaluation. And it works very fine. Tiger's playing sytle has completely changed. Now it plays much safer. Of course it is 30% slower, and that's a bad news because the only real strength of Tiger was the depths it reaches. But I have to do some tuning (lazy eval, and so on), and maybe I'll get a much better program. Christophe
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.