I am a candidate to attend the next a.i. gaming conference in Taiwan this coming September. As such, I have been asked to perform a "peer review" of some of the papers that are being submitted for publishing. One of the papers I will be grading is being written by Dr. Jonathan Schaeffer of the Chinook team.

Dr. Schaeffer has already solved the 10-piece database for the game of checkers. The drawback is, you have to start with 5 kings against 5 kings, then do 5 kings vs. 4 kings + 1 checker, 4 kings + 1 checker vs. 4 kings + 1 checkers, etc., until you get to the most interesing database: 5 checkers vs. 5 checkers.

So, to solve the 11-piece datbase, you need to do 6 kings vs. 5 kings... and the whole mess takes forever.

I proposed a means to "jump right to" 6 checkers vs. 5 checkers, knocking off about 12 years of computation (the databases are really that large) in a casual conversation to his team in 2003. It turns out, I was partially incorrect, but they came up with a means to bridge the gap.

Using a new technique, they are tackling 6 checkers vs. 5 checkers without having to do the tremendously wasteful computation starting with the kings.

It is a fascinating paper.