Author: Bo Persson
Date: 03:14:12 11/25/00
Go up one level in this thread
On November 24, 2000 at 13:16:37, Severi Salminen wrote: >Hi! > >Could someone please explain how Negascout works? Negascout works by searching the first move at each ply with the alpha beta window, and searching all the other moves with an empty window (alpha, alpha+1). The hope is that *all* other moves will fail low. In that case the search is, of course, extremely fast. If some moves returns a score >= alpha, you have to re-search them with the current alpha-beta window to get the real score. If this re-search is rare, and backed by a large transposition table, it will still be a net agin, even though some moves have to be searched twice. >And why is Minimal Window >Search called Principal Variation Search? What does it have got to do with >principal variation? It is the other way around: PVS is a minimal window search. PVS tries to improve on the Negascout by using the best move from a previous search as the first move at each level. Obviously the odds are very high that this move is still the best, and the other moves will quickly fail low. This has been taken to the extreme in the MTD(f) search, where even the first move is searched with a minimal window. You can read about that and also find some interesting links on Aske Plaat's (old) page http://www.cs.vu.nl/~aske/mtdf.html >Severi Bo Persson bop@malmo.mail.telia.com
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.