New Self-Play Results in Computer Chess

  • Ernst A. Heinz
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 2063)


This paper presents the results of a new self-play experiment in computer chess. It is the first such experiment ever to feature search depths beyond 9 plies and thousands of games for every single match. Overall, we executed 24,000 self-play games (3,000 per match) in one “calibration” match and seven “depth X+1⇔X” handicap matches at fixed iteration depths ranging from 5–12 plies. For the experiment to be realistic and independently repeatable, we relied on a state-of-the-art commercial contestant: Fritz 6, one of the strongest modern chess programs available. The main result of our new experiment is that it shows the existence of diminishing returns for additional search in computer chess selfplay by Fritz 6 with 95% statistical confidence. The average rating gain per search iteration shrinks by half from 169 ELO at 6 plies to 84 ELO at 12 plies. The diminishing returns manifest themselves by declining rates of won games and reversely increasing rates of drawn games for the deeper searching program versions. Their rates of lost games, however, remain quite steady for the whole depth range of 5–12 plies.


diminishing returns search vs. knowledge self-play 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Berliner, H. J. and Goetsch, G. and Campbell, M. S. and Ebeling, C. (1990). Measuring the performance potential of chess programs. Artificial Intelligence, Vol. 43, No. 1, pp. 7–21.CrossRefGoogle Scholar
  2. 2.
    Brockington, M. G. (1997). Keyano unplugged-The construction of an Othello program. Technical Report TR 97-05, Department of Computing Science, University of Alberta.Google Scholar
  3. 3.
    Condon, J. H. and Thompson, K. (1983). Belle. Chess Skill in Man and Machine, P.W. Frey (ed.), pp. 82–118, Springer, 2nd ed. 1983, ISBN 0-387-90790-4 / 3-540-90790-4.Google Scholar
  4. 4.
    Gillogly. J. J. (1978). Performance Analysis of the Technology Chess Program. Ph.D. Thesis, Carnegie-Mellon University [printed as Technical Report CMU-CS-78-189, Computer Science Department, Carnegie-Mellon University].Google Scholar
  5. 5.
    Heinhold, J. and Gaede, K.-W. (1964). Ingenieur-Statistik. Oldenbourg, 3rd ed. 1972, ISBN 3-486-31743-1 (in German).Google Scholar
  6. 6.
    Heinz, E.A. (2000). Scalable Search in Computer Chess.Vieweg / Morgan Kaufmann, ISBN 3-528-05732-7.Google Scholar
  7. 7.
    Heinz, E.A. (2000).A New Self-Play Experiment in Computer Chess. Technical Memo No. 608 (MIT-LCS-TM-608), Laboratory for Computer Science, Massachussetts Institute of Technology.Google Scholar
  8. 8.
    Heinz, E.A. (2000). Modeling the “go deep” behaviour of Crafty and DarkThought. Advances in Computer Games 9, Proceedings, H. J. van den Herik and B. Monien (eds.), to be published.Google Scholar
  9. 9.
    Heinz, E.A. (2000). Self-play experiments in computer chess revisited. Advances in Computer Games 9, Proceedings, H. J. van den Herik and B. Monien (eds.), to be published.Google Scholar
  10. 10.
    Heinz, E.A. (1998).DarkThought goes deep. ICCA Journal,Vol. 21, No. 4, pp. 228–244.Google Scholar
  11. 11.
    Hyatt, R. M. and Newborn, M. M. (1997). Crafty goes deep. ICCA Journal,Vol. 20, No. 2, pp. 79–86.Google Scholar
  12. 12.
    Junghanns, A. and Schaeffer, J. (1997). Search versus knowledge in game-playing programs revisited. 15th International Joint Conference on Artificial Intelligence, Proceedings Vol. I, pp. 692–697, Morgan Kaufmann, ISBN 1-558-60480-4.Google Scholar
  13. 13.
    Junghanns, A. and Schaeffer, J. and Brockington, M. and Björnsson, Y. and Marsland, T.A. (1997). Diminishing returns for additional search in chess. Advances in Computer Chess 8, H. J. van den Herik and J.W. H. M. Uiterwijk (eds.), pp. 53–67, University of Maastricht, ISBN 9-062-16234-7.Google Scholar
  14. 14.
    Korf, R. E. (1985). Iterative deepening: An optimal admissible tree search. Artificial Intelligence, Vol.27, No. 1, pp. 97–109.zbMATHCrossRefMathSciNetGoogle Scholar
  15. 15.
    Lee, K.-F. and Mahajan, S. (1990). The development of a world-class Othello program. Artificial Intelligence, Vol. 43, No. 1, pp. 21–36.CrossRefGoogle Scholar
  16. 16.
    Levy, D.N. L. (1997). Crystal balls: The meta-science of prediction in computer chess. ICCA Journal, Vol. 20, No. 2, pp. 71–78.Google Scholar
  17. 17.
    Levy, D.N. L. and Newborn, M. M. (1991). How Computers Play Chess. Computer Science Press, ISBN 0-716-78121-2 / 0-716-78239-1.Google Scholar
  18. 18.
    Levy, D.N. L. (1986). When will brute force programs beat Kasparov? ICCA Journal,Vol. 9, No. 2, pp. 81–86.Google Scholar
  19. 19.
    Moore, D. S. and McCabe, G. P. (1993). Introduction to the Practice of Statistics. W.H. Freyman, 2nd. ed., ISBN 0-716-72250-X.Google Scholar
  20. 20.
    Mysliwietz, P. (1994). Konstruktion und Optimierung von Bewertungsfunktionen beim Schach. Dissertation (Ph.D. Thesis), University of Paderborn.Google Scholar
  21. 21.
    Newborn, M. M. (1985). A hypothesis concerning the strength of chess programs. ICCA Journal, Vol. 8, No. 4, pp. 209–215.Google Scholar
  22. 22.
    Newborn, M. M. (1979). Recent progress in computer chess. Advances in Computers,Vol. 18, pp. 59–117 [reprinted in Computer Games I,D.N. L. Levy (ed.), pp. 226-324, Springer, ISBN 0-387-96496-4 / 3-540-96496-4].Google Scholar
  23. 23.
    Newborn, M. M. (1978). Computer chess: Recent progress and future expectations. 3rd Jerusalem Conference on Information Technology, Proceedings, J. Moneta (ed.), North-Holland, ISBN 0-444-85192-5.Google Scholar
  24. 24.
    Schaeffer, J. (1997). One Jump Ahead: Challenging Human Supremacy in Checkers. Springer, ISBN 0-387-94930-5.Google Scholar
  25. 25.
    Schaeffer, J. and Lu, P. and Szafron, D. and Lake, R. (1993).A re-examination of brute-force search, AAAI Fall Symposium, Proceedings (AAAI Report FS-93-02: Intelligent Games — Planning and Learning), S. Epstein and R. Levinson(eds.), pp. 51–58, AAAI Press, ISBN 0-929-28051-2.Google Scholar
  26. 26.
    Schaeffer, J. (1986). Experiments in Search and Knowledge. Ph.D. Thesis, University of Waterloo [reprinted as Technical Report TR 86-12, Department of Computing Science, University of Alberta].Google Scholar
  27. 27.
    Schubert, F. (2000). Das Ende der Fahnenstange? Über den fallenden Grenzwertnutzen im Computerschach. Computer-Schach & Spiele, Vol. 18, No. 4, pp. 50–54.Google Scholar
  28. 28.
    Slate, D. J. and Atkin, L. R. (1977). Chess 4.5-The Northwestern University chess program. Chess Skill in Man and Machine, P.W. Frey (ed.), pp. 82–118, Springer, 2nd ed.1983, ISBN 0-387-90790-4/3-540-90790-4.Google Scholar
  29. 29.
    Szabo, A. and Szabo, B. (1988). The technology curve revisited. ICCA Journal,Vol. 11, No. 1, pp. 14–20.Google Scholar
  30. 30.
    Thompson, K. (1982). Computer chess strength. Advances in Computer Chess 3, M. R. B. Clarke (ed.), pp. 55–56, Pergamon, ISBN 0-080-26898-6.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2001

Authors and Affiliations

  • Ernst A. Heinz
    • 1
  1. 1.M.I.T. Laboratory for Computer Science (Room NE 43-228)Massachussetts Institute of TechnologyCambridgeUSA

Personalised recommendations