Author: David Rasmussen
Date: 11:19:48 01/10/02
Go up one level in this thread
On January 10, 2002 at 11:56:54, Georg v. Zimmermann wrote: >> >>Mmm. This is a capture extension, not a recapture extension. In many cases, >>capturing the last piece moved _is_ the best move, as the last move was hanging >>a piece. But as I see it, there is no need to extend such lines, because these >>are often idiotic futile lines, that will make a beta cutoff anyway. I think it >>is important to ensure somehow that only captures that brings the value with the >>alpha-beta window are extended. > >What you describe is what everyone is doing. For me, the above works better. > I know. People are not doing the exact same thing, though. I just tried to explain the sensible in doing it like that. >It does extend on lines where a ply ago there might have been an idiotic move, >quite right, but > >a) it will find some combinations much faster A lot of extensions will. The important thing is if it makes your program stronger overall. Seeing a lot of tactics wont necesarily help if you lose a ply in general search, and seeing less positional points. >b) it helps a lot with horizon trouble (thats why your idea with "negative >extensions" doesnt work , it will greatly increase horizon problems) >c) it doesnt make silly lines blow the tree, cause an effective NullMove >Algorithm will catch those. > All sensible extensions help with horizon trouble. The point is: Does it expand the tree too much. And I don't necesarily believe that negative extensions will lead to horizon problems. It has to be tested thoroughly. >Remember that capture extensionsor recap or a hybrid like what I am doing will >in only very rare cases work with a 1 ply extension, I use 1/2 ply. > >Georg Sure. /David
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.