Learning Piece Values Using Temporal Differences

Article type: Research Article

Authors: Beal, D.F.^{; *} | Smith, M.C.^{; *}

Affiliations: London, England

Note: [*] Department of Computer Science, Queen Mary and Westfield College, University of London, Mile End Road, London E1 4NS, England. E-mail: Don.Beal@dcs.qmw.ac.uk.

Abstract: This paper describes experiments where we attempt to learn the relative values of chess pieces by the use of temporal difference learning applied to minimax searches. We show that we are able to learn suitable piece values, and that these values perform at least as well as piece values widely quoted in elementary chess books.

DOI: 10.3233/ICG-1997-20302

Journal: ICGA Journal, vol. 20, no. 3, pp. 147-151, 1997