Author: Uri Blass
Date: 21:46:45 09/18/01
Go up one level in this thread
On September 18, 2001 at 23:08:04, Dave Gomboc wrote: >On September 18, 2001 at 16:33:08, Robert Hyatt wrote: > >>I disagree. If you enter into a series of captures, you might burn all your >>search on the captures and have none left to see what the majority is going to >>do. So you take the first step, and then you are committed. And every move >>you make sees your score drop as you begin to sense the problem. By the time >>the forced trades are done, you realize you are lost, as a king and pawns ending >>can be searched incredibly deeply. > >...which is why the ending should be searched more deeply before the program >actually plays the first capture. GMs don't trade down into king and pawn >endings without expecting a certain result in advance, and neither should >programs. If a program is going to use up its depth, then do a static eval or a >1 or 2-ply search at the beginning of a king and pawn endings, it's just asking >for mistakes. > >Dave If there is a long line then it is usually not a forced line so if the remaining depth in pawn end game is 1 ply there is no point in doing a big search of more than 2 plies and I am not sure even if 2 plies is a good idea because I am afraid it is going to cost too much time. The point is that when the remaining depth is big(for example when the remaining depth is 8 plies and there is transition to pawn endgame programs should search for more than 8 plies(my formula suggested 8*3/2=12 plies) 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.