Author: David Eppstein
Date: 11:07:17 05/15/99
Go up one level in this thread
On May 14, 1999 at 17:11:24, Robert Hyatt wrote: >you would probably want to call it 'gamma'... but it is really both. Either >the score is above 'gamma' or it is below 'gamma'. Below and you fail low, >above and it fails high... very mtd-ish... No, that's not right. I posted the actual code from his thesis in another thread, but basically: Take the usual code for alpha-beta. Only pass in beta as an argument, let alpha be a local variable. Since you don't have a passed-in value for alpha, just set it to -INFINITY at the start of each recursive call. The net effect is that you only get cutoffs from your parent's search, not deep cutoffs. Hsu claims this only loses a factor of 1.07 per level of search.
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.