Author: David Rasmussen
Date: 01:10:06 01/21/02
Go up one level in this thread
On January 20, 2002 at 18:24:33, Uri Blass wrote: >On January 20, 2002 at 16:39:23, David Rasmussen wrote: > >>On January 20, 2002 at 14:36:44, Uri Blass wrote: >> >>> >>>but it is still not clear because the other evaluation stuff is important. >>> >> >>Of course the other evaluation stuff is important. I am not suggesting to cancel >>all evaluation other than material. I am just saying, instead of having a pawn >>be 1 and a knight be 3, and then somewhere in evaluation check if you have >>exchanged a knight for three pawns, then you penalize by, say, 0.5, why not just >>let the knight have a value of 3.5? I know there are more than one requirement, >>which is why it isn't trvial. But there still might be a solution. It is >>basically a linear programming problem. I don't say that material values should >>cover all sorts of evaluation cases, but material _do_ have values right? And >>who says that 1,3,3,5,9 is the Unchangeable Truth? > >I did not say that 1 3 3 5 9 is the unchangable truth. > >I said that the numbers are meaningless without more knowledge about the >evaluation. > >1 3 3 5 9 in a chess program may be eqvivalent to 0.8 3 3 5 9 if you change the >piece square tables. > > >Uri This has nothing to do with piece square tables. I am _only_ taking about material special cases. Three pawns for a bishop. Two minor pieces for a rook and a pawn. Two rooks for a queen. Etc. Everything else will still have to be in the evaluation. The reason for having material values in the first place is _primarily_ to solve such _material_ special cases. Is it good to win a queen for a pawn? Yes, 1,3,3,5,9 says it is. Is it good to win two minor pieces for a rook and a pawn? 1,3,3,5,9 says it doesn't matter. 1,3.5,3.5,5.5,10 says it's a good idea. If the minor pieces had little mobility, and the rook was active it might be a bad idea. Other evaluation terms will measure that. But you will have to have these terms _anyway_. Why have 1,3,3,5,9 and then have code that checks if you have traded two minor pieces for rook and pawn and then give a penalty? /David
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.