Author: Robert Hyatt
Date: 15:08:41 02/04/05
Go up one level in this thread
On February 04, 2005 at 12:51:27, Uri Blass wrote: >On February 04, 2005 at 12:07:34, Robert Hyatt wrote: > >>On February 04, 2005 at 10:50:31, Uri Blass wrote: >> >>>On February 04, 2005 at 10:18:36, Robert Hyatt wrote: >>> >>>>On February 04, 2005 at 06:43:24, S J J wrote: >>>> >>>>>When a program reports a ply count, is it the number of ply that have the moves >>>>>100% generated without pruning, or the deepest node that is checked with >>>>>extensions, or something different? >>>>> >>>>>Steve >>>> >>>> >>>>Depends on the program. Generally accepted practice is to report the number of >>>>plies searched full-width not counting extensions or q-search. Some programs >>>>(such as Junior) use "ply" in a different way which can be misleading if you >>>>compare plies between two programs. >>> >>>Usually it is not full width because of null move pruning. >>>There are a lot of program that use other pruning ideas. >>> >>>Uri >> >> >>Technically null-move is full-width, but reduced depth. I don't throw out _any_ >>moves and not search them, so the search itself is full-width. I just search >>some to reduced depth. > >It is not exactly reduced depth. >If a move has no threat and it is not an endgame you do not search what happen >after legal moves of the opponent so you will never see some zugzwangs. > >You also said: >"Generally accepted practice is to report the number of >plies searched full-width" > >Even without the problem of zugzwangs >12 plies with null move pruning is not 12 plies searched full-width. > >Uri Yes, but the definition of "full width" is to simply generate all legal moves and search them all. MinMax does exactly that. Alpha/beta cheats a bit but does what is called "backward pruning", where some branches are not searched because they have been proven to be irrelevant. "forward pruning" discards moves for whatever reason and they are never searched at all, even if they might have an influence on the final result. null-move doesn't change either of those, it just reduces the overall search depth along branches where there are moves that appear to be pointless... but the "pointless" moves still get searched to some extent...
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.