Author: Uri Blass
Date: 14:56:40 07/02/04
Go up one level in this thread
On July 02, 2004 at 17:22:15, Scott Gasch wrote: >On July 02, 2004 at 12:28:31, Fabien Letouzey wrote: > >>On July 02, 2004 at 12:07:42, Volker Böhm wrote: >> >>>On July 02, 2004 at 09:21:41, Fabien Letouzey wrote: >> >>>>Bruce makes a slightly different assumption that depends on the >>>>window. IMO it is an uncommon implementation but it might be better >>>>depending on aspiration windows. >> >>>What does Bruce suggest? Btw. who is Bruce, Bruce Moreland? >> >>Yes Bruce Moreland. His tutorial is famous, but it seems his description of PVS >>is not the commonly-used algorithm: >> >>http://www.seanet.com/~brucemo/topics/pvs.htm >> >>Fabien. > >Bruce's tutorial doesn't start to search with a minimal window until it find a >PV node (i.e. until a score returned beats alpha). What if you are on, as bruce >calls it, an "alpha node" (where everything is gunna fail low)? If your move >ordering is good then you should still search the best move first. Therefore >you can speed things up even more by searching the rest of the moves with a >minimal window (the thought being that they are all worse than the first move). >Bruce's implementation searches all of these moves on alpha..beta if one has not >yet beat alpha. Maybe it's blasphemy, but I think the "traditional" PVS >implementation is better, assuming you have a good move order. > >Scott I do not understand why you assume that your move ordering is good after a fail low. Suppose that you have a big fail low that means that you cannot save the game after the first move in your move ordering(suppose score of more than 4 pawns against yourself. It means that there are 2 possibilities: 1)You cannot save the game(what you do is not important) 2)You can save the game by a better move(in this case I suspect that it may be better to use bigger window for the other moves in order to save the research for other moves). I do not do it today but it may be a good idea to try it. 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.