Author: Paul Clarke
Date: 05:43:38 11/26/99
Go up one level in this thread
On November 25, 1999 at 09:27:17, Jari Huikari wrote: >Assume that chess program N is playing a game with SLOW time controls. >Number of nodes grows exponentially when search goes iteratively deeper. >Program N thinks that best move is A. However A leads to loss, but behind >the horizon of N. > >It would take terribly much time for N to search deep enough to avoid >the bad move A. So it plays it and looses the game. (Also assume that >N haven't enough knowledge to avoid playing A.) > >PERHAPS it could be avoided to play A, if program N tried to PLAY ON >the game for some moves. If it played on (with shallower search for >each halfmove), it could perhaps see that something very bad happens, if >it plays A. <delurk> I think Chinook (the draughts/checkers program) uses a similar idea. After, for example, an 11-ply search, it plays the first four ply of the PV and then does a 9-ply search of the resulting position. If this shows the score dropping substantially, Chinook will allocate more time and try to search more deeply.
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.