Author: J. Wesley Cleveland
Date: 09:04:09 04/08/03
Go up one level in this thread
On April 07, 2003 at 16:00:32, Dan Andersson wrote: > That would give you a local maximum. To get the minimal tree you would have to >search with *every* possible move order and backtrack when it overflows. If you >then add transpositions the problem gets worse since you would have to reset the >hash when you backtrack. And the hashing scheme would influence the size of the >search. > >MvH Dan Andersson It depends whether you approach it as a mathmatical problem or an engineering problem. To find the true minimum is, as you say, almost impossible for large trees, but if you just want an estimate to see how much your move ordering could be improved, it is easy. At each ply, the minimum number of nodes is 1 + the minimum nodes for the next lower ply for a fail-high, or the number of legal moves + the sum of all the minimum number of nodes for all moves at the next lower ply. A few counters and recursion does this neatly.
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.