A smooth white stone unearthed in the Netherlands has lain in the dark for decades, its purpose debated by archaeologists who suspect it may be part of an ancient board game but have been unable to prove it. In a surprising combination of artificial intelligence and classical archeology, researchers say they have used machine learning to decipher the rules of a mysterious Roman board game. This board game may have entertained legionaries and civilians alike across vast swathes of the Roman Empire.
This breakthrough work announced by a team of scientists represents one of the most novel applications of AI in the humanities to date. Rather than simply cataloging artifacts or translating texts, researchers deployed advanced AI systems to reverse engineer the playable rules of a game that had been forgotten by history for nearly 1,700 years. The discovery shocked both the archaeological and gaming communities and provided a rare window into the leisure activities of ancient people.
A stone that has puzzled generations of scholars
The story begins with a seemingly ordinary artifact. It is a polished white gambling stone recovered from Roman ruins in the Netherlands. As CBS News reported, the stone has long puzzled researchers. Researchers knew it could be a game piece, but lacked contextual evidence to determine which game it belonged to and how that game was played. Ancient board games are notoriously difficult to reconstruct. Although physical components such as boards, stones, and dice may survive for centuries, the rules were usually passed down orally or written on perishable materials that decayed long ago.
Board games were abundant in the Roman world. Ludus latrunculorum, a strategy game sometimes compared to chess, and tabula, the ancestor of backgammon, are among the better-documented examples. However, archaeological sites across Europe have uncovered numerous game boards and pieces that do not neatly correspond to any known game, suggesting that the Romans and their contemporaries played a much wider variety of games than is explicitly shown in the historical record. The Dutch stone seemed to belong to one of these isolated games. The game is a pastime popular enough to produce physical artifacts, but so obscure that no rulebook remains.
Teach a machine a game that no living human can remember
The research team focused on an AI system built on the Ludii General Game System, a digital platform developed as part of the Digital Ludeme project funded by the European Research Council. Ludii is designed to model and play a vast range of board games throughout human history, encoding them as a combination of basic game concepts called “ludems”. This is the basic building block of game mechanics such as capture, movement, stacking, and victory conditions. By assembling Ludem in different configurations, the system can generate and evaluate thousands of virtual games.
According to CBS News, the researchers fed the AI with information about physical evidence, including the size and shape of recovered game boards, the number and type of pieces found at the ruins, and fragmentary historical references. The AI then generated a set of candidate rules that matched this evidence, effectively suggesting a plausible reconstruction of how the game was played. Each candidate was evaluated for playability, strategic depth, and historical relevance, with the machine learning system iterating towards a solution that the human community would find appealing enough to play repeatedly.
Why this matters beyond the game board
The implications of this research extend far beyond recreational curiosity. Board games are first-class cultural artifacts. They encode social values, cognitive priorities, and interpersonal dynamics in a way that pottery shards and cornerstones cannot. A civilization’s games reveal what that civilization thought was fun, fair, and intellectually stimulating. To decipher lost Roman games is, in a meaningful sense, to recover fragments of Roman thought.
Cameron Browne, a professor at Maastricht University and a lead figure in the Digital Ludeme project, has contributed to the development of the computational framework that enables this analysis. Under his guidance, the Ludii system cataloged more than 1,000 traditional games spanning the history of the world, creating what amounts to a computational atlas of human play. The system’s ability to reason about games abstractly (dividing them into modular components and reassembling them) allows it to propose rules for games where only physical evidence survives.
Collaboration between AI and archeology expands
The project is part of a broader trend of deploying artificial intelligence to solve long-standing mysteries in archeology and ancient history. In recent years, AI has been used to decipher damaged ancient Greek inscriptions, reconstruct fragmented Dead Sea Scrolls, and identify previously unrecognized patterns in cave paintings. But the application to board games is unique because it requires the AI to reason not just about what an artifact is, but how it was used. This is a fundamentally more complex reasoning task.
The challenge is similar to finding a deck of cards in an abandoned building 2000 years from now and trying to guess the rules of poker. The cards themselves contain information such as suit, number, and face card, but the leap from physical description to playable game requires an understanding of human psychology, social interaction, and the design principles that make a game satisfying. The Ludii system addresses this problem by incorporating a vast database of known game mechanics, allowing it to identify which combinations of rules are most likely to produce a game that real humans will enjoy.
Skeptics caution against AI reconstruction
Not everyone in the academic community is ready to declare this mystery completely solved. Some academics have warned that AI-generated rulesets, even if plausible, remain hypotheses rather than established facts. Without extant ancient texts that clearly describe the rules, there is no way to verify whether the AI’s reconstruction matches what Roman players actually did. The system identifies the most likely rule given the available evidence, but probabilities are not certain.
Although this caveat is important, it does not detract from the results. In archaeology, certainty is a rare luxury. Scholars routinely conduct probabilistic reconstructions of ancient languages, trade routes, and social structures. The AI-generated game rules meet the same evidentiary standards as many widely accepted archaeological conclusions. These are the best available explanations that match the physical evidence. Furthermore, rules can be tested empirically. Modern players can try out the reconstituted game and evaluate whether it feels like consistent, engaging entertainment or a combination of random mechanics.
What the Romans might have played on a quiet night
Full technical details of the reworked rules have been shared on academic channels, but early descriptions suggest it’s a moderately complex strategy game, one that can be learned quickly but rewards repeated play and tactical thinking. This profile is consistent with what we know about Roman gaming culture, which favored games that could be played with minimal equipment in taverns, military camps, and at home. The game seems to involve moving and capturing pieces on a grid, and the mechanics are somewhat similar to known ancient games, but distinct enough to constitute a separate tradition.
This finding also raises interesting questions about cultural transmission. If this game was played in the Roman Netherlands, a remote province far from the heart of the Mediterranean, it may represent either a game imported by Roman soldiers and administrators, or a local tradition adopted or adapted by the Roman occupiers. Both scenarios deepen our understanding of cultural exchanges along the empire’s northern borders.
A new chapter in digital humanities
Successfully deciphering the rules of this ancient game is a milestone for the Digital Ludeme project and the broader field of computational humanities. This shows that AI can do more than process data at scale. It can perform a kind of creative reasoning and suggest solutions to problems that have resisted human analysis for generations. As the Ludii database continues to grow and the underlying algorithms become more sophisticated, researchers hope to tackle additional missing games from cultures around the world, from Mesopotamian game boards to mysterious African stone games whose rules have never been recorded by colonial-era ethnographers.
For now, the white stone from the Netherlands has finally found its context. After 1,700 years of silence, it is once again part of gaming. Thanks to artificial intelligence, we may finally know how to play.
