Author: Omid David
Date: 14:08:58 09/01/02
Go up one level in this thread
On September 01, 2002 at 12:08:21, Uri Blass wrote: >On September 01, 2002 at 11:55:32, Vincent Diepeveen wrote: > >>On September 01, 2002 at 10:20:08, Uri Blass wrote: >> >>if you search for aske plaat you will find his stuff on >>mtd online probably. i'm amazed that you don't understand >>that Frans is using nullmove. > >I know that he is using null move but I do not use null move when I search the >line that is in the previous pv because I consider it a waste of time. > >Null move is for prunning illogical lines. > >The pv cannot be illogical line so the only case when I can save nodes by null >move pruning when I am in a pv line is when the position is zugzwang. >In other words I can save nodes only if null move pruning is wrong. > >I understood that MTD says that the pv may be wrong but the first ply of the pv >is always right so it does not make sense to prune after Nc6. > >Uri It depends on what you want; if you want "the first move, only the first move, and nothing but the first move", then use MTD(f). But if you also want the PV (as most of us do), avoid it. The use of null-move pruning follows the logic behind each method: In MTD(f), you only want the first move, so you are free to use null-move pruning even in the PV, something you shouldn't do in regular NegaScout/PVS. P.S. Aske Plaat first introduced MTD(f) in his 1995 article "An Algorithm Faster than NegaScout and SSS* in Practice" available at http://www.cs.vu.nl/~aske/Papers/hk.pdf For a more intuitive explanation, look up http://www.cs.vu.nl/~aske/mtdf.html Plaat suggests that MTD(f) is faster than NegaScout, but he researched only fixed-depth full-width trees. I haven't seen any publication concerning MTD(f)'s behavior in variable depth trees (e.g. using null-move pruning).
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.