Author: KarinsDad
Date: 09:01:14 01/20/99
Go up one level in this thread
On January 20, 1999 at 10:08:15, Robert Hyatt wrote: >On January 19, 1999 at 20:20:48, Larry Griffiths wrote: > >>I ran my program with Alpha-Beta and without Alpha-beta. >>The scores and moves chosen were different. >> >>Is this normal for Alpha-Beta or is this an indication of an >>error in my implementation of Alpha-Beta pruning? >> > >It's a bug. Alpha/beta _must_ produce the exact same result as minimax, >this was proven in "An analysis of Alpha/Beta Pruning" written by Knuth and >Moore, 1975.... Actually, it is not necessarily true that there is a bug in the code. There is an assumption on that. Since the min max search should have been exhaustive to a certain depth, you cannot run the program to a deeper depth with the alpha beta or the scores could change which in turn could result in different moves as well. If the alpha beta was able to search deeper than the min max, then there is a bug in the testing as opposed to a bug in the code. KarinsDad > > >> >> >>Also, The time is 1052.58 seconds without Alpha-Beta >>and 3.3 seconds with Alpha-Beta. >> >>What ratio of elapsed time would indicate that I have achieved >>good move ordering etc. with the Alpha-Beta pruning? > >there is no easy answer. If you do no search extensions, and no quiescence >search you could use the alpha/beta 'formula' to compare AB nodes with minimax >nodes and see if the rsult is correct...
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.