How AI Just Cracked a Forgotten Roman Board Game

How AI Just Cracked a Forgotten Roman Board Game

History isn't just written in blood, treaties, and stone monuments. Sometimes, it's etched into a small chunk of limestone by people who were simply bored and wanted to pass the time.

For over a century, a peculiar piece of engraved stone sat in a Dutch museum gathering dust. Unearthed in Heerlen—once the bustling Roman settlement of Coriovallum—the slab featured a grid of lines that didn't match any known game from antiquity. Archaeologists knew it was a game board, but they had no clue how it was played.

Enter artificial intelligence and high-resolution microscopy. Researchers recently managed to decode the rules of this lost Roman pastime, dubbed Ludus Coriovalli, turning an ancient enigma into a playable reality.

The Mystery of the Coriovallum Stone

The artifact itself isn't imposing. Measuring just 21 by 14.5 centimeters, this worked limestone slab was carved roughly 1,500 to 1,700 years ago. When archaeologist Walter Crist first encountered the stone in the Thermenmuseum (now the Roman Museum), he instantly recognized the physical characteristics of a game board.

The problem? The layout—a rectangle featuring four diagonal lines intersecting a straight line—defied standard classification. It didn't look like Ludus Latrunculorum (a Roman tactical game resembling chess) or the ubiquitous Duodecim Scriptis (a distant ancestor of backgammon).

Instead of guessing, the research team looked closer. Literally. Under a microscope, microscopic wear patterns became visible along the incised lines. Glass and stone counters, shoved across the surface centuries ago, left microscopic tracks. Some grooves were a fraction of a millimeter deeper than others, proving intense, repeated friction in specific directions.

Training Algorithms on Antiquity

You can't ask a dead Roman how they moved their pieces. You can, however, write code that simulates every logical possibility until the physical evidence matches reality.

The research team used Ludii, an AI-driven general game system developed at Maastricht University. They fed the computer program the board's exact geometry, combined it with historical knowledge of ancient gaming mechanics, and let virtual agents play thousands of matches using various rulesets.

The algorithms tested roughly 130 different rule configurations—varying how pieces started, how they moved, and what triggered a win. The AI then matched the simulated wear and tear produced by these virtual games against the actual physical abrasion maps scanned from the limestone slab.

One specific rule set created wear patterns that mirrored the museum artifact.

What the AI Uncovered

The results upended traditional timelines for game history. The winning simulation revealed that Ludus Coriovalli was a blocking game.

In this style of play, the goal isn't to capture or eliminate your opponent's pieces, but to trap them so they can't make a legal move. Until now, historians had no concrete evidence of European blocking games dating back to the Roman period; such mechanics were thought to have originated much later, during the Middle Ages.

The most probable version of the game involved asymmetric play: one player maneuvered four pieces against an opponent's two, requiring tactical foresight and defensive mastery.

By bridging high-tech computer simulations with traditional archaeology, researchers have opened a brand new door. We can finally picture ordinary people sitting in the taverns and barracks of Coriovallum, sliding glass beads across limestone, trying desperately to outsmart a friend. Human nature doesn't change much over two millennia; we've always loved finding clever ways to trap each other on a grid.

EM

Emily Martin

An enthusiastic storyteller, Emily Martin captures the human element behind every headline, giving voice to perspectives often overlooked by mainstream media.