Researchers in the Netherlands say artificial intelligence has helped decode how an ancient Roman board game was played. A circular limestone slab unearthed in the country bears intersecting grooves with uneven wear. Using 3D imaging by restoration firm Restaura, archaeologists observed slide marks consistent with moving pieces. An AI system called Ludii—trained on about 100 ancient games from the region—generated candidate rule sets and then self-played to evaluate which were most engaging. The team cross-referenced those with the stone’s wear to propose a likely “blocking” strategy game in which players hunted and trapped opponents’ pieces, possibly made of glass, bone or earthenware. Scholars caution the reconstruction can’t prove how Romans actually played, but say the rules align with comparable games and explain the physical evidence. Findings appear in Antiquity and extend Europe’s evidence for such strategy games back by centuries.
Related articles:
Restoring and attributing ancient texts using deep neural networks
The Vesuvius Challenge: Reading the Herculaneum scrolls with AI
Ludus latrunculorum: a Roman strategy board game
The Royal Game of Ur
