New Results in Deep-Search Behaviour

Article type: Research Article

Authors: Steenhuisen, J. Renze^{; *}

Affiliations: Delft,The Netherlands

Note: [*] Department of Computer Science, Delft University of Technology, The Netherlands, E-mail: J.R.Steenhuisen@tudelft.nl.

Abstract: This article is a follow-up on previous work on deep-search behaviour of chess programs. The program CRAFTY was used to repeat the go-deep experiment on positions taken from previous experiments to push the search horizon to 20 plies. The same experimental setup was used to search, among others, a set of 4,500 positions, from the opening phase, to a depth of 18 plies. Our results showed that the chance of new best moves being discovered decreases exponentially when searching to higher depths, and decreases faster for positions closer to the end of the game. This contribution brings the understanding of deep-search behaviour and the prediction of performance a step further to completion.

DOI: 10.3233/ICG-2005-28402

Journal: ICGA Journal, vol. 28, no. 4, pp. 203-213, 2005