Author: Tom Kerrigan
Date: 13:32:53 04/16/00
Go up one level in this thread
On April 16, 2000 at 15:19:32, Georg v. Zimmermann wrote: >Hi there, > >the program I'm working on uses a simple concept : > >KillerMoves are the moves which are best in the same ply most often. > >History is how often one move has been best in the whole search. > >Is this the most common way to use History/Killers ? Yes. You can do variations on this theme. For example, you can use two killer moves per ply, and you can change the values that you add to your history table. I think most people like adding 2^depth, but you have to be careful of overflows if you do this. >I read in some paper ( I forgot the author sorry ) about SSS and others that the >idea of Killermoves is outdated and almost everyone only uses History today. Is >that correct ? After the history heuristic became popular, I think a lot of people used it to replace the killer heuristic. But there is still some benefit from using both at the same time. My program uses the history heuristic, and it searches about 10% faster with the killer heuristic added. >What is the best way to see how good my move ordering is ? Should I simply count >how often the move "ordered" as number 1 turns out to be the best move ? It might be interesting for you to determine your branching factor. You can do this by counting how many moves you search per node (on average). But the best reflection of your move ordering is simply how efficient your search is. -Tom
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.