Author: José Carlos
Date: 11:05:32 08/03/04
Go up one level in this thread
On August 03, 2004 at 04:54:11, morphy wrote:
>Why the valutation is always referred to the material advantage (where a pawn in
>more is the unit) and we can't have a valutation in tems of percentage of
>probability for winning and drawing?
A chess program scans the search tree (all moves by me -> all replies by my
opponent -> all my replies to all his replies...) until a certain depth (depends
on available time) and evaluates the ending positions (the deepest ones). Then,
it tries to find the best possible move (the one that leads to the better leaf
position by the optimal path) for him. You need a way to score positions and
compare them.
A single number is best choice because comparing numbers is easy and fast. If
you use probabilities, you need one for winning and one for drawing (the one for
losing is 1-w-d). Now you need to compare them to find the best path. How do you
compare w=30 d=40 to w=40 d=20, for example? You can come up with a comparison
algorithm ("maximize chances to win even if I can also lose" and choose second
option or any other idea), but you complicate your code and increase the
probability of mistakes.
Additionally, given a certain position, it's much easier to say "I'm a pawn up
but my opponent has a dangerous attack, worth a half a pawn" and assess
eval=+0.50 than to say "my pawn up has some chances to allow me to win but his
attack is also dangerous" and assess w=40 d=35 (l=25). This example is, really,
much simpler than what a chess program evaluates (piece mobility, center
control, king safety, pawn structure, material, weak squares...). The more terms
you add, the more difficult to use probabilities.
However, using probabilities is not impossible. It would be a nice area to
investigate. I would be happy if someone would come up with an interesting
implementation.
José C.
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.