Author: Robert Hyatt
Date: 11:00:34 03/30/04
Go up one level in this thread
On March 30, 2004 at 03:55:18, Artem Pyatakov wrote: >Hello again, > >Thanks to everyone who has been active in my previous thread - all of these >responses have been helpful and show the diversity of opinion that will >definitely be discussed in my thesis. > >As I attempt to run some experiments for my thesis work, I have run into these >two questions: > >1) Do you think the Bratko-Kopec test suite is at all a reasonable test? Or is >this known to have a lot of problems? Is it outdated? I've been running >experiment on ECM and WAC, but a lot of the previous chess research uses >Bratko-Kopec, so I am thinking of perhaps rerunning my experiments with that >testsuite. > >2) A completely unrelated (to my research) idea has occured to me, as I was >documenting various parts of a standard chess program in the paper. Have people >tried messing around a lot with the Killer Heuristic that everyone finds so >helpful? Do any of you use anything else besides the "2 killers" approach in >Crafty? A very, very quick experiment with about 5 positions showed a measurable >decrease in node counts if I for instance also checked the Killers from two and >four plies lower and from two plies higher. What have people tried in this area? >What about 3 or 4 killers? Chess 4.x used killers from other than the current ply. In Cray Blitz we used all plies we could, sticking with odd/even of course. IE we would try ply N, then N-2, N-4, then N+2, etc. I don't do it in Crafty for simplicity although it is worth another look... I stopped doing it when I added "history", as the two ideas are related... > >Thanks in advance! > >Artem
This page took 0.01 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.