Author: Andrew Platt
Date: 07:33:57 01/18/05
Go up one level in this thread
On January 18, 2005 at 09:44:29, S J J wrote: > > I have a very novice program working without an alpha beta search. On the >surface, it looks like an alpha-beta search can miss a good sacrifice move. > > It does may sense that an alpha-beta search will help speed the evaluation >of a tree of, say, 6 ply. > > However, if there is a sacrifice on the sixth ply that does not gain >the material back until, for example, the 8th ply, won't the node be trimmed >when the sixth ply is evaluated and never have additional moves from that >node generated? I understand your apprehension - I had the same thoughts the first time I was implementing alpha-beta. I had to run many tests with mini-max and alpha-beta to prove (to myself) I was getting the same results. As I really understood alpha-beta I realized that what I was mixing up was the algorithm and the limitations of search-based chess programs. In your example the problem isn't with alpha-beta, it's a fundamental problem that if you search to depth 6 you might find what looks like an excellent move only to find on depth 8 that it loses. That is not a problem with alpha-beta, it's a limitation with search. Read up on quiescent search, SEE, search extensions to see how chess programs try to extend the search down critical lines. Andy.
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.