Dueling AIs reconstruct rules of Roman-era board game.

Researchers used two AI agents to simulate play on a worn oval stone found at Coriovallum and found gameplay patterns consistent with blocking games, supporting the interpretation that the artifact was a game board.

An oval stone recovered at Coriovallum shows a geometric pattern and surface wear that led researchers to propose it functioned as a board game. A study published in the journal Antiquity reports archaeologists used artificial intelligence to test that idea and to model possible rules. The team ran two AI agents through many ancient European game rule sets to see which ones reproduced the board's uneven abrasion. Their simulated play identified patterns consistent with blocking games and strengthened the case that the object was used for gameplay. Key findings: - The artifact’s upper face has a deliberate geometric pattern and wear that is consistent with repeated sliding of game pieces. - Researchers had two AIs simulate play across a large set of small European games, including Haretavl and Gioco dell'orso. - Simulations produced a match with blocking games, where the objective includes limiting an opponent’s movement. - Known traces of blocking games in Europe previously date from the Middle Ages and are rare, so this finding suggests they may have existed earlier. - The authors report this is the first use of AI-driven simulated play alongside archaeological methods to identify a board game. Summary: The study supports the interpretation that the Coriovallum stone was a game board and indicates blocking-style play could predate previously known evidence in Europe. Undetermined at this time.