Crystal Balls

Subtitle: The Meta-Science of Prediction in Computer Chess

Article type: Research Article

Authors: Levy, D.N.L.^{; *}

Affiliations: Londen, England

Note: [*] 89, Constantine Road, London NW32 LP, England. E-mail: [email protected].

Abstract: Deep Blue’s victory over World Chess Champion Garry Kasparov has left us with many questions. Starting with the predictions made in the last two decades, we provide an overview on who was (almost) right and by what margin. Next we compare the educated guesses with two different extrapolations of the increase of the playing strength, viz. based on nodes examined per second, and on the improvement of playing strength per year. The outcomes are guidelines for the answer to the question: “Does the result in New York mean that Deep Blue is now stronger than Kasparov?” Weighing pros and cons, the author arrives at the statement that Deep Blue clearly belongs to the world’s top four, the others being Kasparov, Karpov and Anand. Moreover, Deep Blue’s playing strength is also compared with the best commercial programs. As a conclusion the author assumes that a next match only will take place, if at all, when the title of World Champion is at stake.

DOI: 10.3233/ICG-1997-20203

Journal: ICGA Journal, vol. 20, no. 2, pp. 71-78, 1997