Author: Vincent Diepeveen
Date: 08:20:03 10/02/03
Go up one level in this thread
On October 02, 2003 at 04:51:57, Daniel Clausen wrote: >On October 01, 2003 at 13:54:43, Vincent Diepeveen wrote: > >>On October 01, 2003 at 13:04:18, Anthony Cozzie wrote: >> >>>>consider the search depths diep had last 3 world champs. 9 ply, 10 ply, 11 ply >>>>and where i posted always that you need at least 12-14 ply to get through the >>>>tactical barrier. >>> >>>What do you mean by "tactical barrier" ? >>> >>>anthony >> >>The depth after which evaluation is far more important than search depth. > >Well, that tactical barrier was probably at 12-14 plies because your engine >calculated slightly plies than that. What exactly makes you think there won't be >a tactical barrier at 18-20 plies when your engine calculates 16 plies >typically? It is very obvious, just look to the search depth of the commercial programs. in 1999 i remember junior + fritz getting both 17 ply. Hardware is way faster now but fritz doesn't even get close to 17 in middlegame now. there is plenty of other examples. DIEP in endgame 1999 got around 20 ply at bob's quad xeon. Despite having a very good middlegame evaluation with a lot of knowledge it was clear that endgame needed work. I managed to delay that till end of 2000 and still suffer from a weak endgame in diep i feel (though when compared to some commercial engines it is strong there, take any of the preprocessors endgames) but that's nearly corrected now. Yet search depths are not what they were back then. *everything* gave a cutoff back in 1999. Nullmove really was making extra hours back then :) Yet at 10 ply current version beats the thing from those days being at 20 ply in endgame. So it is very trivial that knowledge is always more important. But trivially WITH that knowledge it is very nice to search as deep as possible. Yet it is not so hard to get 10 ply extra by getting rid of a lot of knowledge. >My redesigned engine typically calculates about 8 plies at the moment and the >tactical barrier seems to be at 9-11 plies. :) > >Sargon
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.