Author: Slater Wold
Date: 09:12:18 02/06/04
Go up one level in this thread
On February 06, 2004 at 12:02:02, Thorsten Czub wrote: >IMO the problem is not the many evaluations you could invent, the problem is >that the SEARCH says: we prune this position because of XYZ reason. >and suddenly you could have the bEST ideas, the program will not see it. > >Take the game CSTAL Shredder, the reason is: why has shredder not seen Bxh6 ? >it was pruned away ! >because shredder thought: it is not working. > >WHY did it think it is not working: because it saw no sense in the moves. >when the moves came on board, and shredder was unable to PRUNE, >it saw the problem. with evaluations ! > >so you should IMO not invent one evaluation after the other, and instead think >how search and score works together such that those cases do not happen. Well of course. If you could solve this, you could probably solve chess. When you do any kind of pruning, it's a trade off. That it will help in more cases than it will hurt. I don't do any kind of pruning or null-move, YET, because of this. I want to build my engine around how it evaluates positions, not around what it will never see. (To explain better, I want a solid engine, with a good search & eval, and doing the things I want/need, and THEN I will try things to speed it up such as reductions & null-move, etc.)
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.