Author: Jay Scott
Date: 13:29:55 09/11/02
Go up one level in this thread
I am fond of best-first style searches for research programs. Always having the whole tree in memory makes things nice and orderly. Both you and the program can understand what's going on more easily, extract statistics after the fact to find out how well things worked, try new ideas with less effort, and all that stuff. If you're going for a "smart" program that applies a ton of knowledge at each node, then the tree will be small and fitting it in memory is not a problem. I even think that most likely it's possible to improve and extend this kind of algorithm until it performs on the same level as today's top programs. But I also think that that would take a lot of new ideas and years of work. It's not the easy path to a winning program. Here they report on an algorithm which they say is better than plain vanilla alpha-beta with no enhancements, and much worse than Crafty. That's exactly what I'd expect from a research report on a new best-first algorithm idea.
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.