A game read from the wear on a stone

For generations, a carved stone in a Dutch museum was a puzzle nobody could solve. In 2026 a team of archaeologists and computer scientists worked out a likely way to play it. They studied where the stone was worn smooth and let computer players try out rules.

The stone

The stone was dug up in Heerlen, in the south of the Netherlands. Heerlen was the Roman town of Coriovallum. The finders were probably working in the late 1800s or early 1900s. The stone went into the town's Roman museum, then called the Thermenmuseum and now Het Romeins Museum. The museum's archives say it was found in Heerlen, and very little else.

Museum number

04433, Het Romeins Museum, Heerlen

Material

White Jurassic limestone, from quarries at Norroy in Lorraine

Size

212 × 145 × 71 mm, 3.38 kg

Likely date

Late Roman, about AD 250–476

The stone was first cut as decoration for a building and was later reused. On its top someone carved a rectangle with a line across the middle and four diagonal lines. Short extra lines sit just outside the rectangle. The pattern looks like a game board, but it matches no board game known anywhere.

Reading the wear

The archaeologist Walter Crist looked at the grooves under a microscope. He saw wear along the lines in the places where you would slide a playing piece. Restaura, a restoration studio in Heerlen, then made very detailed 3D scans of the stone. The scans showed that some grooves are a fraction of a millimetre deeper than others, so some lines were used much more than the rest. Beside one diagonal is a band of stone that is smoother than the surface around it. The band sits 16 to 18 mm from the line, about where fingers holding a sliding piece would rub.

Letting computer players try the rules

The team used Ludii, a game-playing system from the Digital Ludeme Project at Maastricht University. Ludii can run nearly any board game written as a list of rules.

1. They worked out six ways of reading the carved lines as a board. For example, do the diagonals meet the centre line at one point, and are the lines outside the rectangle part of the board?
2. They borrowed rules from games known from other times and places. Most came from haretavl, a Scandinavian game of dogs chasing a hare. There were also line games and blocking games. That made about 130 possible games.
3. Two computer players played each possible game against each other many times. The team counted how often each line on the board was used.
4. They compared those counts with the real wear on the stone. Nine of the 130 games wore the board the same way the stone is worn.

All nine are blocking games: one side tries to hem the other in until it cannot move. Seven of them are four pieces against two, with every piece starting on the board. The authors think Board 6 is the most likely board, and three of the nine games use it. On that board the diagonals meet the centre line at a single point, and the lines outside the rectangle are left out. The authors suggest those outer lines may have been mistakes, which happens on other ancient boards too. This is the "Favoured reading" board in this game.

It was the first time an unknown ancient game had been identified this way.

What we know and what is a guess

Choices this version makes

• You play one round as the dogs and one as the hares. The hares' score is how many moves they make before they are trapped. This follows the scoring in Ludii's rules.
• If the dogs ever have no move, they pass. If the hares reach 40 moves, they have run free and the round ends. Both rules are ours, so that a game can't go on forever.
• We checked both boards with a computer search over every position. With perfect play the dogs can always trap the hares. On the favoured board it takes them at most 7 hare moves when the dogs move first. The "harder" computer plays that perfect game.
• The counters are round, like Roman gaming pieces of glass and bone. Nobody knows what pieces were used on this stone.

Sources

• Crist, W., Piette, É., Jeneson, K., Soemers, D. J. N. J., Stephenson, M., van Goor, … Browne, C. (2026). “Ludus Coriovalli: using artificial intelligence-driven simulations to identify rules for an ancient board game.” Antiquity 100 (409): 111–126. doi.org/10.15184/aqy.2025.10264
• Ludii Portal: Ludus Coriovalli, with the board layouts and the nine scholarly rulesets. This game's boards are taken from Ludii's own rules file.
• Phys.org (Feb 2026): Rules of unknown board game from the Roman period revealed
• Wikipedia: Ludus Coriovalli
• Ludii Portal: Haretavl, the hunt game the rules borrow from.

The stone is on display at Het Romeins Museum in Heerlen.