Author: Cheok Yan Cheng
Date: 00:46:02 06/10/01
Go up one level in this thread
On June 10, 2001 at 02:39:43, Christophe Theron wrote: >The trick is that you do no re-search in the first branches, because the first >branch in the PVS algorithm is searched with the FULL alphabeta window, not a >"null" window (where beta=alpha+1). > >In any position, the first move is searched with an ]alpha;beta[ window, the >returned score becomes the new alpha value (unless there is a cutoff in which >case you quit), and you search the rest of the moves with an ]alpha;alpha+1[ >window. If you get a fail high in any of these moves (and only in this case), >you need to re-search the move with an ]alpha+1;beta[ window. What is the advantages of using ]alpha, alpha+1[ in the 2nd and the rest of the child node? I don't see any point it will produce more cutoff than original alpha beta search. Compare the two windows is ]7, 8[ //Null Window search ]7, 100[ //original alpha beta search If value<=7, ignore the value. (both windows) If value>=100, return the value (both windows) If value>7 AND value<100, re-search arghhhh...... (only Null windows) See! No advantages gain from NULL WINDOWS. This confused me a lot! Please help me with this by providing me some example. >This is a rough explanation of the principle, but I think it highlights a point >you had missed. > >If the move ordering is perfect, you never need to do a re-search, so you have >saved some work because many nodes have been searched with a smaller window. > > > > Christophe
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.