Advertisement

One Jump Ahead pp 253-283 | Cite as

Prelude to Disaster

  • Jonathan Schaeffer

Abstract

With only a few months to go before the Tinsley match I felt that the project was in a state of crisis. Tinsley was right: Chinook hadn’t improved. The success of Tupelo had conned me into overestimating the true strength of the program, denying how lucky we had been, and underestimating how hard it would be to catch up to Tinsley. The Lafferty and Long results had uncovered the true state of affairs: the program was plagued with bugs, and it just wasn’t as good as I would like to have believed. If Tinsley played Chinook today, the result wouldn’t even be close. It was likely to be a crushing victory for the man who had a history of sweeping away all pretenders to the throne.

Keywords

Evaluation Function Opening Book Silicon Graphic World Championship Quick Evaluation 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    In the Software Toolworks tournament Deep Thought tied for first place with grandmaster Tony Miles, ahead of former world champion Mikhail Tal. The program defeated grandmaster Bent Larsen among others. Only former U.S. champion Walter Browne was able to stop the electronic juggernaut.Google Scholar
  2. 2.
    Norman Treloar, personal communication, September 25, 1996.Google Scholar
  3. 3.
    No, the system has nothing to do with Star Trek. The name comes from the model we use to write a parallel program—that of a business organization, or enterprise.Google Scholar
  4. 4.
    Knowing that Herschel had a copy of Chinook, Tinsley once asked to visit him for a few days so he could “prepare” against the program. Of course, Herschel refused—why should Tinsley have access to Chinook when Chinook couldn’t have access to Tinsley. A few months later, Herschel gave Tinsley a ride home from a tournament and, given the long drive, was afraid that Tinsley might try to prod him with questions about his experiences playing Chinook. Instead, Tinsley spent the time discussing religion and trying to convince Herschel to join Marion’s church!Google Scholar
  5. 5.
    Don Lafferty, personal communication, August 19, 1996.Google Scholar
  6. 6.
    Don Lafferty, personal communication, September 4, 1996.Google Scholar
  7. 7.
    I always bring back souvenirs of my travels for my family. While getting dressed on the morning of the third day of the tournament, I noticed a black mark on the middle of my back. I have a mole there and was quite concerned at how large it had become. On arriving back in Edmonton I asked Steph to look at it. The next thing I heard was a scream: “It’s wiggling its legs!” A tick had managed to hitchhike a ride back to Edmonton on my back.Google Scholar
  8. 8.
    E-mail sent on July 27, 1992.Google Scholar
  9. 9.
    There is a famous computer chess game where a similar problem occurred: the COKO III vs. Genie game at the second U.S. Computer Chess Championship in 1971. COKO III had a huge material advantage when it finally found a checkmate in two moves. Genie had nothing else to do except advance an irrelevant pawn. COKO III found a checkmate in two moves, whereas if it had searched a bit longer it would have discovered the mate in one. Genie advanced the same irrelevant pawn. COKO III started searching and found a checkmate in two moves, again stopping the search before finding the mate in one. Genie advanced the same irrelevant pawn. COKO III found a mate in two. Genie advanced the same pawn. COKO III found a mate in two, and so on. Genie’s irrelevant pawn eventually reached the eighth rank, promoted to a queen, stopped the mate threats, and won the game! Had COKO III preferred a mate in one over a mate in two, the game wouldn’t have had such an amusing ending.Google Scholar
  10. 10.
    Several years later, Martin told me what really happened (personal communication, August 7, 1996): “Honest to God Jon there was a bug in the commercial version of Colossus which caused the supposed printer [output] to be dumped on the screen! Fickle finger of fate? Also with a little PC DOS knowledge, if the option had been working, you could have dumped the [output] to a file too! Now that WOULD have been tempting you! ! !”Google Scholar
  11. 11.
    Martin Bryant later wrote (personal communication, August 7, 1996): “I monitored with amusement the outpourings of the EDA during the early Chinook years. At no time did I really consider Colossus the ‘anything’ World Champion (although I may have used the EDA’s decision for commercial reasons upon the odd occasion!) on the basis of it having won a single tournament which was in no way at the time considered a ‘World Championship.’ I was totally realistic about the relative strengths of the programs and would openly admit to anyone who asked that Chinook was almost certainly the stronger overall program. However I believe that I knew probably better than anyone, Tinsley and Schaeffer included, what Chinook’s weaknesses were and how to exploit them. After all, I knew draughts better than Schaeffer and computers bettet than Tinsley! In fact, as I had grown to love the game and thought that it would be bad for the game should Tinsley lose, I would have quite happily helped Tinsley prepare for the match had he asked. Of course, he never did!” Whew!Google Scholar
  12. 12.
    Norman Treloar, personal communication, September 25, 1996.Google Scholar
  13. 13.
    MHz is short for megahertz. The hertz, named after the German physicist Heinrich Hertz, is a measure of frequency, equal to one cycle per second. On many computers, one cycle is equated to executing one machine instruction. The SGI internal clock ticks thirty million times per second. That sounds fast—and it was in 1992—but advancing technology makes this machine look quite slow by today’s standards.Google Scholar
  14. 14.
    ACF Bulletin, October 1992, p. 6.Google Scholar
  15. 15.
    When Lowder swept the pieces from the board, one of them apparently flew to the next table, knocking a checker off the board of a game in progress. Both players were too busy watching Lowder’s antics to notice the missing piece. After the fracas subsided, they continued playing for about forty-five minutes until they noticed that something was wrong with the position. One of the players noticed he was a piece down and asked his opponent, “How did I get to be a piece down?” The opponent replied, “I have no idea!” They asked Richard Fortman, the referee, to intervene. He first requested to see the players’ game records, but neither of them were recording the moves! So he asked them to reconstruct the game. After some effort neither of the players could recall the exact series of moves played. Since nobody could prove anything, Dick made a Solomon-like decision and declared the game a draw. Lowder later wrote a letter of apology to all who were affected by his actions.Google Scholar
  16. 16.
    Gil Dodgen writes (personal correspondence, July 8, 1996): “You had the misfortune of playing your first game against Markusic and drawing an opening that offered him a rare opportunity. Ed had a copy of my program [Checkers] for about a year at that point, and he told me later that he discovered that it didn’t know how to properly play certain bridge endings. He figured that Chinook probably had the same weakness, since these endings are very deep and subtle. Sure enough, he had the chance to lure Chinook into one of these losing bridges. Not only was this a rare coincidence, but had it happened later in the tournament you would have had more [honor] points. Ron King was clearly not the best player at that competition.” A bridge ending is one where White, for example, has checkers on c l and g 1, with a Black checker on e3 creating a “bridge.” The Black bridge allows checkers to crown by going through squares d2 and f2.Google Scholar
  17. 17.
    ; Withrop Lane, “A Realist Looks at Checkers,” Elam’s Checker Board, October 1951, pp. 1673–1676.Google Scholar

Copyright information

© Springer Science+Business Media New York 1997

Authors and Affiliations

  • Jonathan Schaeffer
    • 1
  1. 1.Department of Computing ScienceUniversity of AlbertaEdmontonCanada

Personalised recommendations