Author: Tom Likens
Date: 08:58:00 11/14/02
Go up one level in this thread
On November 14, 2002 at 07:51:31, Vladimir Medvedev wrote: >I want to add mobility terms into my engine's eval function, but it seems to me >rather unclear. > >The primitive way, generating (pseudo-legal) moves in leaf nodes coasts very >much. Checking for move legality will be much more expensive, but without it I >can't evaluate mobiity decrrease for pinned pieces... > >I also tried to evaluate mobility as the last branching factor on the line >before the qsearch begins, but it seems to be too rough estimation, isn't it? > >I think, in bitboard engines mobility can be evaluated in some simple way, but >how to do this for my "field array + piece lists" board structure? If you are satisfied with psuedo-legal moves you could preprocess the non-sliding pieces since their psuedo-mobility will depend only on their location. For example, int KnightMobility[64]; int KingMobility [64]; int WPawnMobility [64] etc. You might also look into incrementally adding mobility to your engine. It seems to be most important for bishops, so you might want to add that first and see how much of an improvement it makes. If you like the results then you can add other pieces. Actually, in general, I'd recommend changing, and then measuring, only one major variable at a time when evalutating new ideas. regards, --tom
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.