Author: Robert Hyatt
Date: 04:25:44 03/11/99
Go up one level in this thread
On March 11, 1999 at 05:10:08, Pauli Misikangas wrote: >Hi! > >I use PVS in my shogi program Shocky. So far I have implemented time controlling >so that when time is up (I pre-define maximum thinking time for each move), I >stop searching immediately and ignore the result from the node I was searching >at the moment and return the previous best one. > >But then I got an idea. If I am re-searching a node on which PVS failed with >null window search, and timeout occurs, should I return this one as the best >node? In other words, if the PVS null window search fails (and return value is >between alpha and beta), do I know that this is the best node so far even >without re-searching the node with wider AB-window? > >Virtually Yours, > >Pauli Misikangas > >PS. Shocky's home page is at http://www.cs.helsinki.fi/~pmturune/Shocky_e.html there is another option. First, on the null-window search I _never_ accept the move when it fails high. I only accept the move if it fails high and _then_ resolves to a higher score, because null-move can cause some erroneous fail highs in odd positions. To solve the problem you mention, I simply don't stop the search until the null-window fail-high is resolved or rejected. IE I treat it like I am losing material and use more time... because I want to be sure I am not rejecting a better move or accepting a worse one.
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.