BKG 9.8: Hans Berliner's Backgammon Program Defeats World Champion Luigi Villa
In July 1979, Hans Berliner of Carnegie Mellon University demonstrated BKG 9.8, a backgammon-playing program, in a money match in Monte Carlo against reigning world champion Luigi Villa, winning 7–1. It was the first computer program to defeat a world champion in a board game under competitive conditions.
Background
By the late 1970s, computers had become reasonably strong at chess, but backgammon was a different kind of problem. Chess has no randomness: both players see everything, and the better position wins. Backgammon has dice. That means the best move on any given turn depends not just on reading the board but on weighing probabilities across many possible futures, and on judging positions that resist simple counting.
The programs that existed before BKG 9.8 could handle the rules and do some basic evaluation, but they were not strong by human standards. The challenge was that backgammon positions are hard to assess without a great deal of specialist knowledge. A program needed to understand concepts like priming (building a wall of consecutive points to trap an opponent’s pieces), blotting strategy, and when to run versus when to fight. Encoding that kind of knowledge in software, without being able to learn it from data the way a human learns from years of play, was genuinely difficult.
Hans Berliner, a computer scientist at Carnegie Mellon University and a former world correspondence chess champion, had spent years thinking carefully about how to evaluate positions in combinatorial games. He believed the key was not raw search depth but the quality of the evaluation function itself: the formula the program used to judge how good a position was.
What happened
Berliner built BKG 9.8 around a set of handcrafted heuristic evaluation functions. Heuristics, in this context, means rules of thumb derived from expert knowledge rather than calculated by searching millions of positions or learned from data. Berliner and his team encoded their understanding of backgammon strategy directly: the program assessed features of the board, weighted them, and chose the move that its evaluation judged best. There were no neural networks involved, and no fuzzy logic. The intelligence was in the carefully constructed rules.
In July 1979, the program played a money match against Luigi Villa, the reigning world backgammon champion, in Monte Carlo. BKG 9.8 won 7–1. Berliner published his account of the match and the program’s design in the journal Artificial Intelligence in 1980, and he was notably candid about one complication: the dice had run in the program’s favour. Villa had not played badly. BKG 9.8 had won partly because it made sound decisions under pressure and partly because the random element of the game had broken its way.
Berliner did not claim the result proved his program was the stronger player in any absolute sense. What the match did show was that a carefully designed evaluation function, built from genuine backgammon understanding, could hold its own against the best human in the world under real competitive conditions, even with the variance that dice introduce. That was new. No computer program had beaten a reigning world champion at a board game before, under conditions that counted.
Why it mattered
BKG 9.8 demonstrated that a rule-based AI system, augmented by carefully crafted evaluation heuristics rather than deep search, could outperform the best human player in a nontrivial combinatorial game. The result raised fundamental questions about whether the program's victory reflected genuine positional understanding or favourable dice variance, a debate Berliner himself engaged with honestly in print, and helped frame later discussions about evaluation quality versus search depth in game AI. It preceded and contextualised subsequent work on learned game evaluation, including Gerald Tesauro's TD-Gammon a decade later.
People
Organisations
Sources
- Backgammon Computer Program Beats World Champion.Artificial Intelligence, Elsevier, vol. 14, no. 2, pp. 205–220.Primary source
- Backgammon Computer Program Beats World Champion (reprint via Backgammon Galore).Backgammon Galore (hosting Berliner's Artificial Intelligence paper).Secondary
Cite this page
AI Achievements. (1979). BKG 9.8: Hans Berliner's Backgammon Program Defeats World Champion Luigi Villa. Retrieved 2026-08-22, from https://achievements.ai/milestone/bkg-program-designed-by-hans-berliner
@misc{achievements_bkg_program_designed_by_hans_berliner,
title = {BKG 9.8: Hans Berliner's Backgammon Program Defeats World Champion Luigi Villa},
author = {{AI Achievements}},
year = {1979},
url = {https://achievements.ai/milestone/bkg-program-designed-by-hans-berliner}
}