Author: Robert Hyatt
Date: 14:05:23 01/28/98
Go up one level in this thread
On January 28, 1998 at 16:50:18, Don Dailey wrote: >Hi, > >I noticed from my discussions with other programmers that there are lots >of small variations of the same thing. > >I have had others claim to use killer captures. I never store a killer >if it's a capture and consider them a separate class. My ordering is >basically the same as Stuart's. Centrality is an improvement I've used >in the past and for some reason have not implemented it. My program >does not respond to the history heuristic for some reason. Even this >algorithm has some variations in how it might be implemented which >should be explored. Do you give each move equal weight or differ based >on distance from root? > >- Don > > from my early testing, you have to weight based on distance from the root... because I trust a move backed up as best near the root rather than one backed up as best near the tips. I use killers, but *only* because I can use them to avoid a move generation at times. They don't make my tree any smaller at all... but the reduce the time per node by about 10% over not using killers. But remember that this is an implementation detail because I have a very efficient way to generate captures without generating the non-captures at the same time... For implementations that can't do this, my idea might not be a win at all...
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.