Author: Robert Hyatt
Date: 07:51:06 06/13/04
Go up one level in this thread
On June 13, 2004 at 10:09:59, S J J wrote: > > I've read that the use of Alpha Beta searches can double the >number of ply a program is evaluating. > On the surface, it looks like both a MinMax and AlphaBeta searches >evaluate the same set of moves that have been generated (Alpha Beta >being able to avoid searching select branches). > To look twice the number of ply deeper, those positions would first >need to be generated. I'd expect generation of the 2X ply to take a >great amount of time compared to the time saved by Alpha Beta. You dont generate moves in positions you don't search. That is why alpha/beta works so well... > > I suspect that "somehow" Alpha Beta enables one to not only avoid >evaluating certain branches, but, also avoid the need to generate them >in the first place. Correct... > > In my program, the amount of time needed to generate the moves is >several times what it takes to evaluate those moves, so I'm assuming >to get the deepest search, Alpha Beta helps trim the number of branches >that generate moves. > Move generation in Crafty is under 10% of the total search time, using a profiler. Evaluation is over 50% of the total time... You just have a ways to go to reach that point. :) > Any clarification would be appreciated.
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.