AI Decodes Rules of Mysterious Roman Board Game After 1,700 Years of Silence

AI Helps Archaeologists Unlock Ancient Gaming Secrets

When Walter Crist walked through a Dutch museum in the summer of 2020, he encountered something that stopped him in his tracks. An oval-shaped limestone tablet, measuring about eight inches across, sat in an exhibit dedicated to the Roman Empire in the Netherlands 2

2

. The ancient Roman game board featured angular lines etched into its surface, forming a pattern that the Leiden University researcher had never seen in archaeological research 1

1

. Dating back 1,500 to 1,700 years to the late Roman Empire, the artifact had been discovered in Heerlen—once the Roman settlement of Coriovallum—sometime in the late 1800s or early 1900s 4

4

. But what game it represented and how people played it remained a complete mystery.

Microscopic Analysis Reveals Critical Clues

Crist, now a guest lecturer at Leiden University, contacted museum curators for closer examination of the stone 2

2

. Working with specialists, his team produced highly detailed 3D scans and conducted microscopic analysis of the 21-by-14.5-centimeter limestone tablet 4

4

. The investigation revealed wear patterns consistent with game pieces—such as glass pebbles—being repeatedly dragged along the surface, particularly along one of the carved lines 1

1

. "Some of those traces are a fraction of a millimetre deeper than others, meaning they were used more intensively," Crist explained 4

4

. The neatly finished edges indicated this was a completed product, not an unfinished stone, strengthening the case that it functioned as an actual game board.

Digital Modeling and AI Simulations Reconstruct Rules of Board Game

Faced with limited historical documentation—the game appeared in no written texts from Roman times—Crist's team turned to artificial intelligence 2

2

. They created a digital model of the pattern and programmed two AI opponents to play against each other using the Ludii game system 3

3

. The researchers tested more than 100 different sets of rules taken from other known European games, both ancient and modern, including Scandinavia's Haretavl and Italy's Gioco dell'orso 5

5

. As the AI agents played 1,000 games per rule set, the team tracked how pieces moved and compared the gameplay patterns with the levels of wear on areas of the board 2

2

.

Blocking Game Discovery Rewrites European Gaming History

The AI simulations showed that wear patterns were most consistent with a blocking game—one in which the goal is to prevent the opponent from moving their pieces 1

1

. The team identified nine rule sets that appeared consistent with the board's wear, and "they were all variations of this same kind of blocking game," Crist noted 2

2

. The results suggest one player took turns placing four pieces in the grooves against an opponent's two, with victory going to whoever avoided being blocked the longest 3

3

. The researchers named it Ludus Coriovalli, Latin for "the game from Coriovallum," and the game can now be played online 2

2

.

Groundbreaking Method Opens New Research Possibilities

This Roman board game represents the earliest example of such a blocking game in Europe—a type previously thought to have emerged only during the Middle Ages 1

1

. "This is the first time that AI-driven simulated play has been used in concert with archaeological methods to identify a board game," Crist stated 5

5

. Véronique Dasen of Switzerland's University of Fribourg called the study "groundbreaking" and noted the technique could investigate other lost games 3

3

. The research results invite archaeologists to reconsider Roman period graffiti that could represent actual boards for similar games not present in texts. Published in the journal Antiquity, this approach provides archaeologists with tools to identify games from ancient cultures that are unusual or uncommonly played, since current identification methods rely on connecting geometric patterns to games known from historical references or artistic representations 5

5

.