Author: Robert Hyatt
Date: 16:44:44 11/24/97
Go up one level in this thread
On November 24, 1997 at 19:24:55, Willie Wood wrote: >All, > >I recall reading that some programs, like Ferret, don't use mobility in >their eval functions, and I would guess that's for speed. I've been >trying to do without mobility as an experiment, and it's been a bit >tough to get the bishop to know what it's doing. Anybody have a hint? > I'm not sure what Bruce does. I do mobility for bishops, as there is not a lot of other things that can tell whether the bishop is a good one or a bad one... But it doesn't have to be an expensive calculation, if you use the right data structures. IE mobility is "free" for bitmap programs... >Another issue is the tree explosion that occurs with check extensions. >Hyatt mentioned the use of fractional ply extensions on replys-to-check, >though I don't see how that avoids being arbitrary (my lack of >understanding). I wonder if there's a good check-pruning method that >would avoid chasing the king all over the board? > >WW I have used fractional ply extensions for about 15 years. The idea is that a normal check can extend 1 ply, but if there is only one legal way out, I let this extend by 3/4 of a ply. The 3/4 is critical, because a single such extension does nothing (3/4 < 1). but the second such extension will trigger an extra ply, as will the third and fourth, then the fifth won't extend, but the next 3 will. This provides a throttling effect to prevent a non-terminating search. Just be aware that following too many checks is just as bad as following too few. You want to find the obvious mates so you don't get killed, but there is no real need to find a mate in 30. I've played in computer vs computer and computer vs human games for almost 30 years now and have *never* seen a program of mine announce a mate in 30, nor have I ever found a position where such was important. If you go too far, you waste time that could be used in other ways...
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.