Author: Alvaro Jose Povoa Cardoso
Date: 12:03:20 01/05/02
Go up one level in this thread
>Can someone explain the difference between history and killers? > >john coffey Killers are somewhat 'ply dependant'. That means that if a move was good at say ply=3 at some branch of the tree, it will probably be good at ply=3 at another branch (if it is legal). History moves are more generally and they mean that if a move was good at some node it will probably be good at another node of the tree regardeless of the ply at witch they are. For the moment my experience is only in checkers, and in my program (Profound) killers don't give me much, but history moves really helps. I think that killers are more helpful in chess because chess has a much greater branching factor so a killer move has much higher probability of causing a fail high. Also in chess there are frequentely more equally good moves at a given node. In the portuguese version of checkers (with flying kings), most of the time there is only one good move (I see this because very frequentely my program locks on a move and never changes his mind). I must say however that if I use a plain alphabeta searcher with no hash tables and no history moves then killers can help something. Regards, Alvaro Cardoso
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.