Author: Uri Blass
Date: 13:37:17 08/02/04
Go up one level in this thread
On August 02, 2004 at 16:22:14, Gian-Carlo Pascutto wrote: >On August 02, 2004 at 16:12:46, Uri Blass wrote: > >>On August 02, 2004 at 16:04:14, Gian-Carlo Pascutto wrote: >> >>>On August 02, 2004 at 16:01:14, Cesar Contreras wrote: >>> >>>>Hi >>>> >>>>I'm getting better result's (about 8% margin) ordering capture moves based only >>>>on victim value (ignoring attacker value), i think i'm doing something grong. >>>> >>>>My question is: it's significatly better MVV\LVA over MVV >>>> >>>>I know both are aproximations, but witch one it's statically better. >>>> >>>>Thanks in advance. >>> >>>Did you use attacker values instead of attacker *indexes*? >>> >>>It's a common mistake and I can't think of anything else that would cause what >>>you're seeing. >>> >>>-- >>>GCP >> >>I am afraid that I have no attacker indexes in my code so I do not understand >>what you are talking about. >> >>I have piece list but it is only an array like bishops[10][2] that gives me the >>squares of the bishops in the board(no more than 10 bishops per side). > > MVV/LVA value MVV/LVA index >Pawn 100 1 >Bishop 300 2 >Knight 300 2 >Rook 500 3 >Queen 900 4 > >Sortvalue = Victimvalue - Attackerindex > >-- >GCP I understand I use something that is probably slightly better than MVV/LVA but I start with the following formula (piece(to) * 8) - piece(from) piece(PAWN)=0 piece(KNIGHT)=1 piece(BISHOP)=2 piece(ROOK)=3 piece(QUEEN)=4 piece(KING)=5 I later use my attack tables to correct this formula. Note that the formula always mean that capturing bigger piece is better and when the piece that is captured is equal it prefers smaller attacker. It may be better to have piece(KNIGHT)=piece(BISHOP) and I did not test it. Uri
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.