In the 1970s, in a grave in a Bronze Age cemetery in Shahr-i Sokhta, Iran, an incredible object was unearthed next to a human skull: the oldest complete board game ever discovered. Around 4500 years old, it consists of a board with 20 circular spaces created from the coils of a carved snake, four dice and 27 geometric pieces.
The Shahr-i Sokhta game is one of many ancient board games discovered around the world, such as the Roman game Ludus Latrunculorum and the Egyptian game Senet, found in Tutankhamun’s tomb. But we have only been able to guess how to play these games. There are no preserved rulebooks – with the notable exception of the Royal Game of Ur from ancient Mesopotamia, whose long-lost rules were deciphered in 2007 from a cuneiform tablet in the British Museum.
Advertisement
Now, though, another tool is helping to bring these games back to life. In recent years, researchers have been harnessing artificial intelligence to assist in the hunt for likely rules. The goal is to make these forgotten games realistically playable again, while also gaining insights into the evolution of game types. “These games act as a window into the past, offering glimpses into the social and cultural dynamics of the people who played them,” says at the Catholic University of Louvain in Belgium.
Board games were an important part of life in the ancient world. “Game boards are found all over the world from antiquity,” says at the British Museum, who discovered the . For other ancient games, we could only make educated guesses as to how they might have been played, based on scant historical records, if available, and what we know from current games about what makes them fun to play.
Universal rulesets?
Another complication with establishing plausible game rules is that ancient gamers were unlikely to follow a single universal ruleset. “There were no game boxes that you could buy in a shop together with the rules,” says at the University of Fribourg in Switzerland. “Somebody had to explain to you how the game worked.”
This oral teaching process probably spawned multiple ruleset variants as games spread across different geographic regions over hundreds or even thousands of years, says Schädler. But the upside of this is that, today, we don’t need to be obsessed with finding the definitive rules for the game, we just need to get close to plausible ones.
Schädler is among those who have already begun deducing such rulesets for Ludus Latrunculorum, also known as the Game of Brigands (or Mercenaries), based on strategies used in games played today plus archaeological and historical evidence, like a Latin poem that praises the skill of Roman aristocrat Calpurnius Piso. The game was played on a stone or ceramic board divided into a grid of squares, with game pieces probably consisting of glass discs, and featured two players trying to surround and capture opponent pieces.
A similar approach has allowed researchers to propose rulesets for many other ancient games, including those for Senet and the Game of Five Lines, played by ancient Greeks. These have been brought to life through online gaming sites like and the , which let you play out the rule variants proposed by researchers such as Finkel and Schädler on digital boards, with pieces and dice rendered through simple visuals.
The experience is a far cry from the immersive computer-generated graphics of the latest video games – but my board gamer’s heart fluttered more than once at the sight of my digital opponents making their ghostly moves in an echo of ancient players from past millennia. Software developer Sam Jelveh and archaeologist Hossein Morad have also created a , with a charmingly digitised wooden board, using based on those of the Royal Game of Ur, which it closely resembles.
Now AI and computer simulations are further enhancing the reconstruction of these games. One of the goals is that, in future, if a game like the one found in Shahr-i Shokhta is unearthed, an AI would be able to suggest methods for how it was actually played. “One key method is AI-driven rule generation, where algorithms simulate various plausible rulesets based on the game’s structure,” says Piette.
Another way that AI can help is testing how the myriad permutations of possible rules play out, to find out which are fun and which lead to tedium. This is done by breaking the game down into units of game-playing information and feeding these “ludemes” into an AI.
One of the first case studies for this AI approach has been Ludus Latrunculorum – one of the ancient games that we know the most about because of historical writings. “This gave our reconstruction process the best chance of success,” says at Maastricht University in the Netherlands, who led the Digital Ludeme Project (DLP) – a five-year project that ran until 2023 and that “investigated the full range of over 1000 traditional games throughout 6000 years of human history”.
Historical descriptions point to Ludus Latrunculorum having been played on rectangular or square boards, but archaeologists have discovered boards varying in size from six-by-seven to 17-by-18 squares at sites scattered across the former Roman Empire. This raises the question of whether the game was played on all these boards or whether some boards were used for other games.
Browne and his colleagues in the DLP used AI-simulated gameplay to investigate which board sizes make the most sense for proposed rules of the game. He worked with at Leiden University in the Netherlands and others to identify three games still played today – Kharebga, Seega and Tablut – with rulesets bearing a strong resemblance to Ludus Latrunculorum. They adapted the rules to accommodate the different-sized boards. The three rulesets were then played by opposing AI gamers.
These simulations showed that game sessions became mind-numbingly long as the board size increased, leading the researchers to conclude that the smaller boards were most suitable to the descriptions of Ludus Latrunculorum found in ancient Roman sources. “The larger boards, which are found on the outskirts of the Roman Empire in some of the forts in Spain, Egypt and Britain, are probably some other games that we don’t know about,” says Crist. Exactly which games is a mystery yet to be solved.
Another achievement of the DLP was to create a , documenting their rules and historical contexts, and highlighting cultural links across different regions and periods.
Game evolution
Looking forward, more than 200 computer scientists, archaeologists and historians have teamed up in a new European network called to develop even more sophisticated AI-powered digital tools for studying historical games. One goal is to create more human-like AI players. Another is to extend the DLP games database.
Using this kind of family tree, the evolution and origins of games can be traced back over time. Through this, we might discover that many ancient games didn’t die out, but developed into games still played today. Finkel, for example, thinks that the Royal Game of Ur was ultimately replaced by backgammon. “A ‘good’ board game… is unlikely to die,” he says, “but will evolve into something more vital.”
Article amended on 13 December 2024
This feature was amended to acknowledge Cameron Browne’s role as leader of the Digital Ludeme Project, and also the games database created by this project
