|
What if we applied Dr. Eliyahu M. Goldratt's Theory of Constraints and Thinking Process to a computer chess
program?
Theory of Constraints is a thinking process that requires people to make logical decisions based on the current environment
using key barometers... the barometer must correctly model the system and reflect the constrained activities of the organization.[D.
Smith, The Measurement Nightmare, p.77]
The common theme running through all TOC applications is constraint management. Because constraints
are what keep an enterprise from reaching its goal, global optimization of enterprises has to address constraints. [John Ricketts,
Reaching the Goal, p.63]
Wikipedia article on Theory of Constraints
Read more about the Theory of Constraints
This interesting idea for an evaluation function for a computer chess program is worth reading if you enjoy looking
at new ideas in the field of computer chess. We use the future mobility of the chess pieces and the ability of the lower-value
enemy pieces to restrict mobility to better estimate positional pressure and focus our search efforts. Some of these
ideas have previously been mentioned by Dan Heisman (in a slightly different form) in Elements of Positional Evaluation.
This paper can be considered to be an extension of the ideas of Dan Heisman, M. Botvinnik, Eliyahu M. Goldratt, Judea
Pearl, and Aron Katsenelinboigen.
Abstract: How might we create an evaluation function for a computer chess
program that plays a stronger positional game of chess? A new heuristic for estimating the positional pressure produced
by chess pieces is proposed. We calculate and maintain a database of "potential mobility" for each chess piece 3 moves into the future, for each
position we evaluate in our search tree. We update this piece
mobility database dynamically as we evaluate each new chess position (this database allows us to reward chess pieces for specific objectives they can accomplish in the
future). We determine the restrictions placed on the future mobility of the pieces based on the mobility of the lower-valued
enemy pieces. The central idea is that a better search focus is achieved by this method, and that the projected slower speed
of the evaluation function is compensated by a realistic opportunity for selective search.
Initial results are presented.
A condensed version of this paper, and recommended starting point:
http://mysite.verizon.net/vzesz4a6/current/jerz.pdf
|
