Author: Frank Schneider
Date: 09:46:16 05/31/00
Go up one level in this thread
On May 31, 2000 at 12:24:20, Roberto Waldteufel wrote: >On May 31, 2000 at 12:01:22, Ed Schröder wrote: > >>On May 31, 2000 at 04:16:44, Roberto Waldteufel wrote: >> >>>I am experimenting with some move ordering heuristics and I would like to know >>>for comparison what percentage of moves proposed and searched by the killer >>>and/or history heuristic lead to cut-offs in other programs. Many thanks in >>>advance for any info >>> >>>Roberto >> >>It depends. It will strongly depend on what you already have in your >>program. If you (for instance) don't have "the best move from the hash >>table" in your program the "killer heuristic" will give you a lot more >>than if you already have the "best move from the hash table". >> >>But implement a 2-slot "killer heuristic" by all means. It should give >>you a clear speed-up. Maybe just 5-10%, maybe a lot more. As already >>said, it depends. >> >>Ed > >Hi Ed, > >Thank you for my response. Perhaps I should have made myself more clear - what I >am doing is experimenting with a new heuristic that suggests a move to try for a >quick cut-off. I wanted to know how effective it was compared to the commonly >used methods for doing this. I do try the hash table move first, then the move >proposed by my test heuristic if the hash move did not produce a cut-off. My >heuristic generates a cut-off about 65%-75% of all the times it is invoked. It >is meant to be an alternative to history/killer heuristic for closing move >prediction, but I am not sure how well this figure compares to the well tried >methods. I measured how often the first move causes a cutoff in fail-high-nodes and got 92% (the same number was reported for crafty some time ago). My moveordering is hashmove first, then good captures. I also experimented with different kinds of killers, but there is not much room for improvement... Frank > >Best wishes, > >Roberto
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.