AI Reconstructs 1,800-Year-Old Roman Board Game Ludus Coriovalli
Ancient Roman Board Game
I explore how researchers used AI to revive Ludus Coriovalli, an ancient Roman game lost for over 1,800 years. By analyzing wear patterns on a limestone board found in Heerlen, the team reconstructed the rules for this asymmetric strategy game where four Hounds block two Hares. This landmark study proves that microscopic analysis and AI simulations can bring forgotten history back to life.
"It is not a game that was passed down through centuries of unbroken tradition; rather, it was lost to history for over 1,800 years until artificial intelligence reconstructed its rules from physical traces."
HN discussion
- The study utilized the Ludii General Game System to run Alpha-Beta search agents across candidate rulesets, matching simulated edge-usage statistics against physical wear patterns on the artifact to identify nine plausible configurations.
- Skeptics argue that archaeology's reliance on 'best guess' reconstructions and the introduction of AI uncertainty creates a narrative that prioritizes media-friendly crisp stories over acknowledging the vast missing pieces of the puzzle.
- Critics contend that the AI's preference for a four-versus-two asymmetric game may reflect a modern bias toward balance, whereas ancient designers might have intentionally encoded asymmetry to mirror natural hierarchies or master-slave dynamics.
- Defenders of the methodology emphasize that such reconstructions are standard in archaeology, where conclusions are explicitly framed as best estimates based on current evidence and are subject to revision as new data emerges.