Author: Don Dailey
Date: 18:22:02 09/21/98
Go up one level in this thread
On September 20, 1998 at 19:31:36, John Coffey wrote: >We all know that searching 1 ply, 2ply, 3ply, 4ply etc produces faster >results than just searching 4 ply by itself because the move odering >improves the efficiency of the alpha-beta pruning. > >But ... > >Has this approach ever been used above the base of the tree? i.e. If >I am doing a seven ply search and I make the first move for white and >I am searching the black responses, at this point would it pay to do >a 1ply, 2ply, 3ply, 4ply, 5 ply, and finally a 6 ply search? > >If so, then the number of sub-searches would get quite high. > >John Coffey So what you are talking about is called rercursive iterative deepening. It was used in a program called star-tech which I did not write. If you have good move ordering heuristics, this probably does not pay off. I have a tiny piece of code in my program that simply does a reduced depth search if a hash table suggestion is not available, just to get a best move. It is a tiny improvement in most positions, occasionally a respectable one. Most of these ideas are minor refinements. I think every chess programmers is looking for that one line code change that will add 100 rating points! I doubt it exists (unless it is a bug fix!) - Don
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.