Author: Robert Hyatt
Date: 15:58:57 12/28/05
Go up one level in this thread
On December 28, 2005 at 12:46:30, Tord Romstad wrote: >On December 28, 2005 at 12:31:13, Stuart Cracraft wrote: > >>Hi - I ran a small experiment last night. >> >>Many people don't put mobility into their evaluation function because of the >>heavy expense of a move-gen at the terminal node. > >I am fairly sure almost everybody uses some kind of mobility in their >evaluation function. The only exception I know among reasonably strong >programs is Crafty (I am not quite sure, though -- I haven't looked at >Crafty recently). yes and no. Crafty does some mobility. bishops have a normal mobility term, for example, but with rotated bitboards, the cost is almost nothing. The first eval I wrote was 100% mobility. But many things are commonly used to replace mobility with an indirect form. Rooks on open files for example. Increases mobility, but that mobility is better than a rook that attacks more squares,but in a limited way. I don't do any sort of knight mobility, and once even did queen mobility although I do not today. There are _lots_ of ways to get to mobility without actually doing mobility. "knight on the rim is dim" is a measure of mobility, or the lack thereof, for example. > >As long as you only consider pseudo-legal mobility, it isn't expensive >at all. There is no need to do a move-gen at terminal nodes; it is enough >to count the number of squares each piece attacks. > >If you want to calculate safe mobility (i.e. the number of moves each piece >can make without losing material), achieving good performance is more >challenging. I don't know any very efficient ways to do this without using >attack tables. > >Tord
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.