Chinook becomes the first program to win the right to play for a human world championship

Background

Chinook began in 1989 as a project aimed at building a program that could defeat the human World Checkers Champion ^[1]. Schaeffer has said he moved into checkers work in 1989 because the game offered the same research challenges as chess while being less complex ^[2]. Early copies of his program started running in 1989 on undergraduate machines at the University of Alberta ^[2]. Checkers may look simple, but the project page gives its search space as 5 times 10 to the 20th power ^[1]. A search space on that scale means a program cannot simply examine every possible line by hand; it needs search methods and evaluation to choose moves. That is why the project was a computing research effort as well as a game-playing project.

The 1990 milestone

The central event came in 1990, when Chinook became the first program in any game to win the right to play for a human World Championship ^[1]. This was a qualification result, not the championship match itself. The program had earned a place against the human champion, making it the first computer program to reach that stage in any game, according to the project page ^[1]. For the project, the result turned a research program into a participant in the human championship structure. It also gave the team a direct test: how would a program built for checkers perform in a formal match against the human champion?

Championship results

The project page records that Chinook lost the Championship match in 1992 and then became Champion in 1994 ^[1]. Those two results frame the 1990 qualification. The first championship match did not produce a win, but the program later reached the championship level. The project did not stop at match play, however. Dozens of computers worked almost continuously since 1989 to solve the game, with a gap from 1997 to 2001 ^[1]. That long-running computation was aimed at a different kind of result: not just beating a particular opponent, but determining the game's outcome under perfect play.

Solving checkers

On April 29, 2007, the team announced that checkers was solved, with perfect play guaranteeing a draw from the standard starting position ^[1]. The announcement completed a line of work that had begun in 1989 and had already produced a human championship qualification and a championship title. In a solved game, the result is known in advance when both sides play perfectly; for standard checkers, that result is a draw ^[1]. The 1990 milestone therefore sits between the project's original goal of defeating the human champion and its later goal of solving the game itself.

Why the milestone matters

For AI, the 1990 result showed that a program could qualify for a human world championship in a game with a search space of 5 times 10 to the 20th power ^[1]. It also marked the start of a sequence: qualification in 1990, a lost championship match in 1992, a championship in 1994, and a solved game announced in 2007 ^[1]. The later work did not make the 1990 result unimportant; it showed what the qualification opened up. Chinook's first championship shot was the point at which computer checkers entered the human championship system, and the project's subsequent history turned that entry into a longer research program.