A limestone slab excavated in 1984 in Heerlen, on the site of the Roman settlement Coriovallum, drew attention because of its incised grooves and marks. Although humans have played board and strategy games since at least the Bronze Age, most informal game equipment (sticks, pebbles, grooves scratched on the ground) seldom survives archaeologically, so objects that could be game boards are rare and significant. For the Heerlen find, no surviving rulebook or close match among known Roman or earlier European games existed to confirm its function.
Previous microscopic wear analyses showed uneven abrasion along the carved lines: some tracks were much more worn than others, suggesting repeated, patterned movements rather than random decoration. Still, researchers lacked a convincing gameplay model that would reproduce the observed wear pattern.
New approach: 3D scans and artificial intelligence
A multinational research team led by Maastricht University, with Leiden University, Flinders University, the Louvain Catholic University (Belgium) and the Restaura Roman Museum and restoration studio in Heerlen, adopted a different approach: they produced high-resolution 3D scans of the limestone and used AI-driven simulations to test possible rule sets.
The idea was straightforward: a game played many times should leave characteristic wear traces. Two AI-controlled agents were programmed to play on a digital replica of the carved board, using more than 100 different rule systems drawn from historically known games and their variants.
For each rule set the AI players ran 1,000 simulated games. The researchers then compared the distribution of simulated move frequencies to the microscopic wear pattern on the stone surface.
Results: match with blocking-style games
The simulations revealed a strong correspondence between the observed wear and the patterns produced by simulations of so-called blocking games: two-player strategic games in which the objective is not to remove the opponent but to restrict their movement. In these simulations, movement concentrated along the same carved lines that show the strongest wear on the limestone.
Other rule systems tested did not reproduce the same wear distribution. Based on these results, the researchers concluded that the limestone likely functioned as a game board for a two-player blocking game.
Significance for archaeology and cultural history
The findings support the interpretation of the object as a Roman-period game board and push back evidence for the existence and play of blocking-type games in Europe by several centuries. This sheds new light on Roman strategic pastimes and the sophistication of their game culture.
Walter Crist, the project lead, summarized: “This is the first time that AI-driven simulated play has been used together with archaeological methods to identify a board game. This research also provides tools for archaeologists to identify unusual or rarely played games from ancient cultures, given that current identification methods rely on geometric patterns of a playing surface and their links to games known today.”
The team notes that, while absolute certainty about the exact historical rules is impossible, the consistency of the simulation results makes it feasible to reconstruct playable rules and thereby partially recreate the historical experience.
Methodological advance
Beyond this specific object, the study demonstrates a reproducible method combining 3D digitalization, quantitative microscopic wear analysis, and AI-based gameplay simulation. This approach gives archaeologists a new way to identify and test hypotheses about game objects lacking clear documentary or iconographic parallels.
(The limestone piece was excavated in Heerlen in 1984; the institutions involved in the project reported the simulation parameters and core findings.)
