Backgammon Bots History
As Play65 continues its efforts to get rid of backgammon bots, it is a good opportunity to look back at the history of the computer programs that often play better than the best human backgammon players, for better and for worse.
The first backgammon bot that succeed in defeating a championship level player, and not just a player but the then World Backgammon Champion, Luigi Villa of Italy, was BKG 9.8. It was developed in the late 1970s by the computer science professor and World Correspondence Chess Champion Hans Berliner with the assistant of former World Backgammon Champion Paul Magriel.
The human vs. machine match took place in June 1979, right after Mr. Villa was crowned the new world champion in Monte Carlo. They played 5 games and the final score was 7-1. Prof. Berliner earned the $5000 prize on behalf of the bot, but later justified the program’s victory with the lucky appearances of better rolls than those rolled by the opponent, who all – including Berliner - agreed played far better than the machine.
The next step in backgammon bots history was taken at the end of the following decade with the publication of the first commercial bot, Expert Backgammon. Although our robotic backgammon expert skill level fell from this of a human backgammon expert, the latter benefited from the former greatest novelty – the ability to perform rollouts (analyze positions by playing them out many times) persistently.
Thus, Expert Backgammon made a greater impact on the game then BKG 9.8, allowing players in analyzing different positions and finding out the best possible play in any given position, and making backgammon a more statistically proven game, what has been theoretical when rollouts where performed manually by human beings. Backgammon computer revolution was completed with TD-Gammon, the first neural net backgammon program that equaled a champion level player. And this story worth a whole different post. Soon.
