How wear on a Roman game board helped researchers narrow down its rules

The surviving evidence was a stone board, not a set of instructions. For researchers studying an object from Heerlen in the Netherlands, its incised lines suggested places for pieces to move. Uneven wear suggested something more: those pieces may have travelled some routes more often than others.[2]

The study, published in Antiquity on 11 February 2026, combined close examination of the surface with simulations in the Ludii game system. Nine rule-and-board configurations met the researchers’ criteria. All were blocking games: the smaller side would try to escape being immobilised by the larger side.[1]

Turning wear into a question a computer can test

A playable reconstruction has several unknowns. How many pieces did each player have? Where did they start? Which lines could they follow? What counted as a win? A drawing alone leaves room for many entertaining answers. The wear supplies a further constraint: a candidate game ought to produce movement that is compatible with the stone’s uneven surface.

The team used surface-recording techniques including photogrammetry and photometric stereo to examine wear. It then tested candidate games with alpha–beta search agents, allowing one second per move and running 1,000 rounds for each board-and-rule combination. The criteria included especially frequent use of a diagonal and lower use of other routes.[1]

Alpha–beta search is a way of examining possible future moves while discarding branches that cannot improve the current choice. Here, the AI’s job was to play proposed games consistently enough to generate movement patterns. It was not reading a hidden Roman rulebook from a photograph.

The outcome was a family of games

Those unequal forces are the interesting result. A blocking game can be competitive without giving both sides the same number of pieces, because their objectives differ. One side needs to restrict movement; the other needs to keep an escape available.

The journal’s account places the interpretation within the history of blocking games and explains why the apparently simple object became more informative when simulation was added. The computer did not contribute a new excavation find. It made it practical to compare the consequences of many possible reconstructions.[2]

Why a plausible reconstruction can still be valuable

Maastricht University’s account emphasises that simulations can identify playable rules without making the historical reconstruction certain. The surviving game proposals depend on the candidate rules, the agents’ behaviour and the interpretation of the wear. A different plausible search space could contain further answers.[3]

An everyday analogy helps. Suppose an old garden has several stone paths and one is much more worn. Repeated journeys between a gate and a shed could explain the pattern. So could a different routine that uses much of the same route. The wear narrows the stories; it does not narrate them by itself.

For the game board, the same reasoning turns uncertainty into useful questions. Would alternative starting positions reproduce the heavily used route? Would less skilful play change the comparison? Would another surviving board with similar markings support the same family of games? These are potential follow-up tests, not claims of results already obtained.

The advance is a reconstruction that can be played, measured and challenged against a physical object. Instead of stopping at “this looks like a board”, the researchers could ask which kinds of play would leave traces resembling the ones that survived.

Report an error or suggest a correction