A 2,000-year-old Roman puzzle solved by AI — and it’s a game

It’s not every day that a weathered slab of stone sitting quietly in a small Dutch museum becomes the crossroads of archaeology, computer science and human curiosity. Yet Object 04433 — an unassuming piece of white Jurassic limestone from the Roman site of Coriovallum (modern Heerlen) — has just had its story rewritten. After more than a century of head-scratching, high-resolution scanning, use-wear forensics and simulated play by AI, researchers now argue the slab was a playable board: a variant of a “blocking” game related to haretavl (hare-and-hounds) traditions. The team calls the reconstructed game Ludus Coriovalli.

Why this matters goes beyond one artifact. The study shows how digital tools can recreate behaviors lost to time, turning scratches and smoothed lines into a living rule set. That kind of detective work — mixing microscope-level physical evidence with millions of simulated moves — is archaeology at its most 21st-century.

Quick takeaways from the discovery

• The object (Object 04433) is a rounded, intentionally shaped slab of Norroy limestone found at Coriovallum, now in the Het Romeins Museum collection in Heerlen, Netherlands.
• Microscopic and photogrammetric analysis revealed uneven wear: some incised lines are noticeably more abraded, consistent with repeated gameplay along those tracks.
• Researchers ran 1,100 AI-driven simulated games across 130 rule configurations (using the Ludii platform) and found that blocking-style games best reproduce the observed wear pattern.
• If correct, this pushes back evidence for blockade-type games into the Roman period in northern Europe, suggesting a deeper and older distribution for this family of games than previously known.
• The project highlights a new method for identifying ancient play: combine use-wear analysis, 3D imaging and AI simulation to infer plausible rulesets from material traces.

The object and the problem

At first glance, Object 04433 looks like a roughly rectangular block — but closer inspection shows deliberate shaping, bevels and an engraved network of lines not seen on typical Roman building stones. For decades scholars debated its purpose: a decorative piece, an architectural plan, a tile fragment — or a game board.

The breakthrough came when researchers treated the slab as an archaeological palimpsest of behavior. Using photogrammetry and photometric stereo, they generated precise 3D depth maps that made subtle wear visible. Certain lines had been smoothed by repeated abrasion; others remained sharp. That unevenness is the fingerprint of repeated human action, not random erosion.

How AI helped turn scratches into rules

This is where the study gets clever. The team didn’t just compare the slab to known board geometries; they built candidate games from rules documented across northern Europe, then used Ludii — a formal game-description and simulation system — to run thousands of AI-played matches for each ruleset. The idea: if players repeatedly use certain tracks during play, those lines should show higher simulated usage and thus match the wear observed on the stone.

After testing hundreds of permutations (different piece counts, movement and capture rules, starting positions, and so forth), the AI simulations that most closely matched the wear patterns were variants of blocking or pursuit-and-encirclement games — think “hare and hounds” and related traditions. In short: the stone likely hosted games where one side tried to trap the other, producing repeated movement along particular lines.

What this reveals about Roman life

• Play as routine: Finding a dedicated object for a relatively local or regional game suggests structured leisure — not just impromptu play in the dirt. People invested time and materials into play.
• Cultural overlap: The reconstructed rules link Roman-period material culture to game forms known from later medieval and northern European sources, revealing deep continuities or diffusion channels for certain game types.
• Methodological shift: This study offers a template for reading behavior from artifacts that initially seem inscrutable. Wear patterns + AI-driven behavior modeling = plausible reconstructions of how ancient people lived and played.

Wider implications and limits

There’s an alluring simplicity to the idea that AI “decoded” an ancient board game, but the real advance is methodological: pairing rigorous surface analysis with simulated behavior. The authors are careful — the match is strong but not unique. Alternative explanations (manufacturing marks, non-game uses, post-depositional processes) can’t be absolutely ruled out. Still, the convergence of physical evidence and simulation makes the gaming interpretation persuasive.

This approach also raises exciting possibilities. Museums and archaeologists hold countless objects whose purpose is unclear; many might reveal human practices if examined with the same forensic and computational toolkit. At the same time, we should remember that AI doesn’t conjure facts out of thin air — it amplifies hypotheses and tests them against measurable traces. Human judgment, comparative knowledge and archaeological context remain essential.

My take

There’s something charming about connecting a two-millennia-old pastime to the same human impulse that fuels modern board-game nights. That this connection was revealed by AI underscores how technology can deepen — not replace — our understanding of the past. The slab doesn’t just become an artifact with a label; it regains part of the life it once hosted: bodies leaning over a table, fingers nudging pieces, laughter, stakes, perhaps even wagers. That kind of bridging between eras is the best of archaeology.

Sources

• Ludus Coriovalli: using artificial intelligence-driven simulations to identify rules for an ancient board game — Antiquity. https://www.cambridge.org/core/journals/antiquity/article/ludus-coriovalli-using-artificial-intelligencedriven-simulations-to-identify-rules-for-an-ancient-board-game/ (DOI: https://doi.org/10.15184/aqy.2025.10264)

• A Mysterious Roman Artifact Covered in Odd Markings Puzzled Researchers for More Than a Century—Now AI Has Cracked Its Ancient Code — The Debrief. https://thedebrief.org/a-mysterious-roman-artifact-covered-in-odd-markings-puzzled-researchers-for-more-than-a-century-now-ai-has-cracked-its-ancient-code/

• Experts use AI to crack mystery of 2,000-year-old Roman board game found in Netherlands — Yahoo News (re-reporting). https://www.yahoo.com/news/articles/experts-ai-crack-mystery-2-220239777.html

• High-tech strategy reveals secret behind Roman game of stones — The Times (reporting on the study).