Author: Severi Salminen
Date: 00:28:24 11/27/00
Go up one level in this thread
On November 27, 2000 at 03:15:30, Bruce Moreland wrote: >On November 27, 2000 at 02:57:37, Gian-Carlo Pascutto wrote: > >>On November 26, 2000 at 13:18:52, Bruce Moreland wrote: >> >>>1) MVV/LVA. You take the *value* (pawn=100, minor=300, etc.) of the taken >>>piece, and subtract the *index* (pawn=0, knight=1, bishop=2, rook=3, etc.) of >>>the attacking piece, and sort this from high to low, and just search them all. >> >>I always thought MVV/LVA was taking the value of the taken piece and subtracting >>the value of the taking piece. I've never heard of it this way. Why would >>it work better? > >It is a case of primary/secondary key. The primary key is the value of the >taken piece. The secondary key is the value of the taking piece. > >This is how GnuChess does it. I assumed that it was doing MVV/LVA correctly. In Heinz's book "Scalable Search in Computer Chess" he defines the order of MVV/LVA: PxQ, NxQ,...,RxP, Q,P. So this is indeed the right way to perform this kind of sorting. If we subtracted the value of attacking piece instead of the index we would have PxN sorted near NxR which should be higher. But I have not tested which one is better in practice. Have to test. Severi
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.