One Jump Ahead pp 103-126 | Cite as

The Case for the Prosecution

  • Jonathan Schaeffer

Abstract

January 1990 was a busy month. Getting back into teaching mode was difficult after eight months of fun—pardon me—research. And, of course, there were the hectic preparations for my wedding. Steph and I have spent many wonderful holidays in Jasper in the Rocky Mountains. One of the magic places to visit is Maligne Canyon in the winter. In the summer, it’s a deep gorge with a raging river. In the winter, everything is frozen, so it’s possible to go down to the bottom of the canyon. On the frozen waterway, fifty meters below the ground, you can see spectacular ice formations, frozen waterfalls, and large rooms that have been created by centuries of erosion. It was here in the middle of winter that Steph and I were married. A small group of immediate family and close friends attended. We returned directly to Edmonton the next day so I wouldn’t miss a lecture. That meant a postponed honeymoon, something my wife reminds me of even today.

Keywords

Evaluation Function Capture Move World Championship Strong Player Prize Money 
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.
    Ken Thompson wrote software to read the Encyclopedia of Chess Openings. He too found many errors in this book. See: Henry Baird and Ken Thompson, “Reading Chess,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 12, no. 6(1990): pp. 552–559.CrossRefGoogle Scholar
  2. 2.
    Gerry Tesauro has developed a backgammon program that is better than Berliner’s BKG 9.8 and is now playing at the level of the top human players. It is based on a learning technique called neural nets. The learning is considerably more sophisticated than what we tried in Chinook and, apparently, a lot more successful. See: “Temporal Difference Learning and TD-Gammon,” Communications of the ACM, March 1995, pp. 58–68.Google Scholar
  3. 3.
    E-mail sent on November 9, 1994.Google Scholar
  4. 4.
    He also played for the world eleven-man ballot championship later in 1990, losing to Elbert Lowder. Many players argue that the three-move ballot is played out; the good players know all the drawing lines. In eleven-man ballot, one piece is randomly removed from the front two ranks on each side and then two moves are randomly made. This gives rise to over two thousand unique openings, forcing both players to rely exclusively on their playing skills, without the benefit of extensive opening knowledge.Google Scholar
  5. 5.
    To represent five unique values of win, loss, or draw requires 3* 3* 3* 3* 3= 35 = 243 values. A byte has eight bits, representing 28 = 256 values. There isn’t enough “room” in a byte to store more values.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