The quest to build a machine that could play checkers (draughts) better than a human is one of the foundational stories of artificial intelligence. Long before Deep Blue defeated Garry Kasparov in chess, computer scientists were using checkers as the ultimate testing ground for machine learning and search algorithms.
The story begins in the 1950s with Arthur Samuel, an IBM researcher who wrote one of the world's first successful self-learning programs. His checkers playing program, developed on an IBM 701, was groundbreaking because it was one of the first demonstrations of a computer learning from its own experience.
Samuel's program played thousands of games against itself, adjusting the weights of its evaluation parameters based on the outcomes. By 1962, the program had achieved a level of play strong enough to defeat a respectable human player, marking a major milestone in the public perception of AI.
While Samuel proved machines could learn, they were still far from beating world champions. Fast forward to 1989, when a team led by Jonathan Schaeffer at the University of Alberta began working on "Chinook". Their goal was singular: win the human World Checkers Championship.
Chinook combined deep search algorithms (alpha-beta pruning) with a massive endgame database. This database contained perfect play for all board positions with eight or fewer pieces. If a game ever reached a state with eight pieces left, Chinook stopped "thinking" and simply looked up the guaranteed optimal path to a win or a draw.
Chinook's greatest rival was Dr. Marion Tinsley, widely considered the greatest checkers player who ever lived. In their first major match in 1992, Tinsley defeated Chinook, proving that human intuition still held an edge over raw computational power.
However, by 1994, Chinook had been significantly upgraded. In their rematch, the two played to six consecutive draws before Tinsley sadly had to withdraw due to health issues (he passed away shortly after). Chinook became the official World Champion, the first time a computer had ever won a human world championship in any game.
Jonathan Schaeffer wasn't satisfied with just winning the championship. He wanted to mathematically prove the game. After 18 years of computational effort, peaking at 50 computers running continuously, the team announced in 2007 that Checkers had been weakly solved.
The conclusion? Perfect play by both sides always leads to a draw. No matter what opening you choose, if both the Red and Black players make the optimal mathematical move every turn, neither side can force a win.
Today, the algorithms pioneered by Samuel and Schaeffer run effortlessly in web browsers. Tools like Next Checkers Move bring this immense computational history to your fingertips. Modern devices can instantly calculate deep into the game tree, providing optimal moves and evaluating complex positions in milliseconds.
While the game may be "solved" mathematically, it remains a rich, challenging, and deeply entertaining game for humans, with the AI now serving as an incredible training partner.