> No written rules for this game survived antiquity. To reconstruct how the game may have been played, researchers turned to the Ludii General Game System — a comprehensive digital platform developed at Maastricht University that can model and simulate thousands of historic board games. The results were published in the journal Antiquity (Volume 100, Issue 409, 2025).
> Using Alpha-Beta search agents — the same class of algorithm that powered early chess computers — the team ran 1,000 simulated rounds for each candidate ruleset, allowing one second of processing time per move. The AI tracked which lines on the board were used most frequently during play, generating detailed edge-usage statistics....
> Nine game configurations matched the wear criteria. All of them were blocking games, and the most frequently matching format was a four-versus-two game in which pieces start on the board. This site faithfully reproduces one of these AI-validated configurations.
I would say this is more "inspired by" Ancient Rome.
This feels like the thing that makes me deeply skeptical of swaths of archaeology and palaeontology as a science rather than being a kind of fandom.
I imagine the incentives of having a crisp story for media consumption don’t help. I’d hope to read a lot more: “we’re missing the majority of the pieces to this puzzle. This represents our best guess given current evidence and methods.”
Archaeology is an unusual discipline in that it incorporates so many others as tools. Chemistry, physics, geology, CPSC, you name it. It's difficult enough to figure out what people were doing based on ruins and trash pits. It's harder still when there are so many disciplines involved that, each, introduce their own uncertainties.
That being said, "We asked an AI..." is a special kind of uncertainty that goes above and beyond anything else Archaeologists do.
--------
"No written rules for this game survived antiquity. To reconstruct how the game may have been played, researchers turned to the Ludii General Game System — a comprehensive digital platform developed at Maastricht University that can model and simulate thousands of historic board games. The results were published in the journal Antiquity (Volume 100, Issue 409, 2025).
Using Alpha-Beta search agents — the same class of algorithm that powered early chess computers — the team ran 1,000 simulated rounds for each candidate ruleset, allowing one second of processing time per move. The AI tracked which lines on the board were used most frequently during play, generating detailed edge-usage statistics.
These statistics were then compared to the physical wear patterns on Object 04433. To account for human cognitive biases — such as right-handed players preferring to play on the right side of the board — the researchers applied symmetry transformations to the simulation results, maximising consistency between AI-generated play and the actual marks left by ancient players.
Nine game configurations matched the wear criteria. All of them were blocking games, and the most frequently matching format was a four-versus-two game in which pieces start on the board. This site faithfully reproduces one of these AI-validated configurations."
--------
It's interesting that they considered use-wear on found pieces as input for their AI. Still, this study made a lot of assumptions. I wouldn't be surprised if a different team could use the same methods and come up with a completely different result.
I think the study would be more palatable if it was presented as an exploration of AI-aided methods and very strongly impressed that any one result was not the point.
There's a huge difference between "asking and AI" meaning a prompt to a LLM, and very particular simulation fine tuned to this problem where the evaluation criteria is matching the exact wear found in the real boards.
This work is much closer to a numerical simulation than anything you can call AI.
Their AI is "alpha-beta search agents" which, unless that's a very misleading description, sounds like a bunch of instances of minimax with alpha-beta pruning [1]. Which indeed should be the simplest, cheapest and most sensible option for that kind of game. I mean, no need to run an LLM/ agent harness on such a simple board game, that'd be overkill. And, well, tbh, it wouldn't work as well.
So they could have used wear patterns as a component in an evaluation function used in minimax but I get the feeling they most likely did a wear analysis afterwards to weed out some of the possible rulesets indicated by the minimax play patterns.
The actual paper is written by archaeologists for archaeologists - just like all papers are written by people in a feild for people in a field.
Your complaint is like complaining that a CS paper doesn't even mention that P = NP is still unknown, and that it just assumes that the best sorting methods are O(n log n). Some things are considered general knowledge in a field, and in archaeology one of those things is: all of this is a best guess given current evidence and methods.
The whole paper is full of citations to other work, which is full of citations to other work - all this work linked includes detailed reports of what was found where and what-else was there, people who do statistical analyisis of similar findings, people who ask questions like "what explanations can there be for it?", more importantly "what else would we find if X was true, if Y was true?". When new evidence arrives, people can and do go back and re-examine these things. Your random skepticism and all the questions you may ask have already been asked and addressed, and frankly: these archaeologist's conclusions carry far far far more believable weight than your half-assed skepticism.
True, this is the best we can get and archaeology is full of speculation.
It is still a very, very reasonable speculation, probably 95% accurate.
You can thank the ones who burned the Biblioteca of Alexandria for most of the loss of actual information.
> No written rules for this game survived antiquity… a comprehensive digital platform… that can model and simulate thousands of historic board games.
Model and simulate based on what?
> Nine game configurations matched the wear criteria.
So their idea was to generate candidate rulesets, have AI try to figure out rational play, then see which pieces would be moved most often and match that to the forensic evidence?
This seems impossible. If you started with chess pieces and a chessboard or even a go board and go pieces, it is absolutely impossible to reconstruct the game as they’re currently played.
Whichis is fair and good, bc nobody would expect you to rebuild the actual rules of chess from the board alone.
However it would still be useful if archeologists used the board to figure out some games similar to checkers, or go; or if they also have the pieces they could guess it was a combat game like Shogi. Any of those would give you insight about the kinds of leisure that people may get from that board.
The thing missing here is that nearly all of the games that use these nodes + stones are a variant on Nine Men's Morris (or Twelve Men's Morris, or etc). Discovering the ruleset for a particular variant is kind of like discovering a new version of "Uno" based on uncovering new house rules.
Even if the researchers did not uncover the exact right set of rules, it would probably not be dissimilar from an actual variant that could have been played somewhere in Rome.
I really liked this game, but it frustrates me a little bit that there only seems to be one way to win. Or at least I've only found one. If I play the hardest AI, I can consistently win, with the right sequence, but it always ends up on the right side with the hares next to each other.
Is it like chess only way more obvious, that the first mover wins, and if so, wouldn't the Romans have figured it out?
Or am I just exploiting a weakness in the AI and a human would make better choices?
I was thinking of them yesterday because I noticed one for sale in the antiquities shop in Andor, which brought up all sorts of Earth/Rome/Star Wars cannon questions.
The best explanation anyone has come up with for those dodecahedrons was as tools for making jewelry from gold wire. Here's a video by an expert knitter named Amy Gaines in which she demonstrates the process: https://www.youtube.com/watch?v=lADTLozKm0I
The segment where she uses a replica dodecahedron to make a chain from aluminum wire (stiffer than gold, meaning this would have been easier with actual gold) is entirely convincing. It explains everything about the shape of the objects, both the protruding knobs on the end and also the fact that the holes are of different sizes.
There is a decade old paper that compiled over 50 possible explanations, and many more have been hypthesized since. Some of the dodecahedrons are solid, which may negate many of the theorized uses.
I always like to hear about ancient board game reconstructions. Like music and religion, games are something every culture creates. Another recent example is the case of Liubu, a game from ancient China for which the rules were lost. Still, a reconstruction is being attempted by Carnegie Melon:
Maybe the concept of desiring balance is modern. Maybe they wanted to encode asymmetry in the game to mirror nature, or the Master / Slave dynamic which they likely considered to be a natural law.
The symmetry is restored if you say you're playing a two game match and the winner is the one that survives longer as the hares. But the game seems simple enough that I imagine it shouldn't be too hard to play optimally.
I won second try with hares on medium, I think the trick is to:
A) make sure you get at least one hare to one of the central two spots (will always be possible) and camp it there. Keep your other hare moving around on your side.
B) the opponent will have to bring two hounds over to catch your other hare. To do this they will have to create a gap between the first and second hound in the middle spot above/below your camped hare. As soon as this happens, move your camped hare upwards.
It's pretty easy after that TBH. The AI has usual burnt a load of moves by that point and you have so many options from that position
I beat both on hard fairly easily? Came back to comments expecting remarks on the shape of the board allowing for fairly easy "I just move back an forth as a hare" and "I just wait for you to move wrong as a hound"
As hares, if you rush one of the middle nodes, the hound machine could capture, but it never figured out moving two hounds all the way over to do so in my play. As hounds, it's basically the same, you just make sure none of the hares ever get "behind" you because then it's a lot harder to contain them.
Since the rules are entirely reconstructed, I have to say I'm a bit skeptical of a game where a central point is keeping a count of moves up to 150. It seems too unpractical for a casual game.
Are there other known ancient games that work like this?
i think i broke the bot! i won on hard w the hares.
figured out that my hares needed to stay together in either the top or bottom center and left corners of the big square.
at one point the bot started wasting moves so i repeated mine until it ran out! not sure if I found the deterministic win of the game or it was a bug in the AI.
The AI could have constructed any number of rules seeing as how it's literally a "best guess" and it chose rules that aren't very fun. Congratulations, I guess.
Edit: Or, put a different way: Sometimes the rules to games are lost for a reason; they're not very good.
Similar in that it is asymmetrical, very old, and we don't quite know what the rules were (though with tafl we have slightly better historical evidence, including an account from Carl Linnaeus in 1732, which has allowed us to produce a few educated guesses). In fact tafl is sometimes speculated to derive from the Roman game of ludus latrunculorum - I'm not sure if that is the same game described here.
Think of this as a very old game, with a board simple enough you can trace it on the ground and play with a few stones, with rules simple enough you can explain/understand in 30 seconds. Like tic-tac-toe, mancala, nim, chopsticks... simplicity and portability wins over perfect rules or balance.
The videos on the game (and all his other videos) with Irving Finkel, a curator at the British museum, are spellbinding. He has the looks, manners and enthusiasm of an eccentric museum curator from central casting!
> A partial description in cuneiform of the rules of the Game of Ur as played in the second century BC has been preserved on a Babylonian clay tablet written by the scribe Itti-Marduk-balāṭu.
This is much more palatable and cool since they’re not just randomly guessing what the game can be like this article is
It's far from random. Reconstructing it with game playing AIs is a bit excessive maybe, but blocking games (hare games) are a style of game which are very widely historically attested.
They are not randomly guessing, they are doing something more sofisticated: they are correlating gameplay with the evidence of board wear.
They are creating a probability distribution function where each point is a different gameplay ruleset and then we are free to pick the gameplay ruleset with the highest probability, with the caveat that there are other similar gameplays, with slightly different rules, and almost the same probability. And there are other gameplay rules with much lower probability because they don't match the wear of the board.
A random guess would have a flat distribution where any gameplay has the same probability.
The rules seem like a rather boring pre-decessor to the mills game/nine men's morris. I guess this is what I would draw inspiration from though, to make the game more interesting, because the board is actually kind of neat.
> No written rules for this game survived antiquity. To reconstruct how the game may have been played, researchers turned to the Ludii General Game System — a comprehensive digital platform developed at Maastricht University that can model and simulate thousands of historic board games. The results were published in the journal Antiquity (Volume 100, Issue 409, 2025).
> Using Alpha-Beta search agents — the same class of algorithm that powered early chess computers — the team ran 1,000 simulated rounds for each candidate ruleset, allowing one second of processing time per move. The AI tracked which lines on the board were used most frequently during play, generating detailed edge-usage statistics....
> Nine game configurations matched the wear criteria. All of them were blocking games, and the most frequently matching format was a four-versus-two game in which pieces start on the board. This site faithfully reproduces one of these AI-validated configurations.
I would say this is more "inspired by" Ancient Rome.
This feels like the thing that makes me deeply skeptical of swaths of archaeology and palaeontology as a science rather than being a kind of fandom.
I imagine the incentives of having a crisp story for media consumption don’t help. I’d hope to read a lot more: “we’re missing the majority of the pieces to this puzzle. This represents our best guess given current evidence and methods.”
I have often wondered how to make this clear wrt science communication
Eg much is not known about dinosaurs. Many things cannot be found in fossils (obligatory xkcd: https://xkcd.com/1747/)
How do you communicate what is unknown or what can't be known?
My favorite dinosaur theory is they looked far more like a plump turkey than a Komodo dragon.
We tend to just drape a bounding skin volume over the bones and give it a color, MAYBE some feathers.
But a whale skeleton looks nothing like the mass of flesh of a whale. Skeletons give no clue about an Elephant’s head shape.
So imagine a huge fluffy owl or yellow chick with a trex skeleton deep inside.
Archaeology is an unusual discipline in that it incorporates so many others as tools. Chemistry, physics, geology, CPSC, you name it. It's difficult enough to figure out what people were doing based on ruins and trash pits. It's harder still when there are so many disciplines involved that, each, introduce their own uncertainties.
That being said, "We asked an AI..." is a special kind of uncertainty that goes above and beyond anything else Archaeologists do.
--------
"No written rules for this game survived antiquity. To reconstruct how the game may have been played, researchers turned to the Ludii General Game System — a comprehensive digital platform developed at Maastricht University that can model and simulate thousands of historic board games. The results were published in the journal Antiquity (Volume 100, Issue 409, 2025).
Using Alpha-Beta search agents — the same class of algorithm that powered early chess computers — the team ran 1,000 simulated rounds for each candidate ruleset, allowing one second of processing time per move. The AI tracked which lines on the board were used most frequently during play, generating detailed edge-usage statistics.
These statistics were then compared to the physical wear patterns on Object 04433. To account for human cognitive biases — such as right-handed players preferring to play on the right side of the board — the researchers applied symmetry transformations to the simulation results, maximising consistency between AI-generated play and the actual marks left by ancient players.
Nine game configurations matched the wear criteria. All of them were blocking games, and the most frequently matching format was a four-versus-two game in which pieces start on the board. This site faithfully reproduces one of these AI-validated configurations."
--------
It's interesting that they considered use-wear on found pieces as input for their AI. Still, this study made a lot of assumptions. I wouldn't be surprised if a different team could use the same methods and come up with a completely different result.
I think the study would be more palatable if it was presented as an exploration of AI-aided methods and very strongly impressed that any one result was not the point.
There's a huge difference between "asking and AI" meaning a prompt to a LLM, and very particular simulation fine tuned to this problem where the evaluation criteria is matching the exact wear found in the real boards. This work is much closer to a numerical simulation than anything you can call AI.
The wear on the pieces and board was not an input for the AI tool they used.
Also, they didn’t “ask” an AI anything, as it was not a natural language AI tool.
Their AI is "alpha-beta search agents" which, unless that's a very misleading description, sounds like a bunch of instances of minimax with alpha-beta pruning [1]. Which indeed should be the simplest, cheapest and most sensible option for that kind of game. I mean, no need to run an LLM/ agent harness on such a simple board game, that'd be overkill. And, well, tbh, it wouldn't work as well.
So they could have used wear patterns as a component in an evaluation function used in minimax but I get the feeling they most likely did a wear analysis afterwards to weed out some of the possible rulesets indicated by the minimax play patterns.
______________
[1] https://en.wikipedia.org/wiki/Alpha%E2%80%93beta_pruning
The actual paper is written by archaeologists for archaeologists - just like all papers are written by people in a feild for people in a field.
Your complaint is like complaining that a CS paper doesn't even mention that P = NP is still unknown, and that it just assumes that the best sorting methods are O(n log n). Some things are considered general knowledge in a field, and in archaeology one of those things is: all of this is a best guess given current evidence and methods.
The whole paper is full of citations to other work, which is full of citations to other work - all this work linked includes detailed reports of what was found where and what-else was there, people who do statistical analyisis of similar findings, people who ask questions like "what explanations can there be for it?", more importantly "what else would we find if X was true, if Y was true?". When new evidence arrives, people can and do go back and re-examine these things. Your random skepticism and all the questions you may ask have already been asked and addressed, and frankly: these archaeologist's conclusions carry far far far more believable weight than your half-assed skepticism.
What about reading the paper before you proclaim all archeology rotten over the wording in a wikipedia article?
New here? This is the best place to find Dunning Kruger in the wild!
True, this is the best we can get and archaeology is full of speculation. It is still a very, very reasonable speculation, probably 95% accurate. You can thank the ones who burned the Biblioteca of Alexandria for most of the loss of actual information.
> No written rules for this game survived antiquity… a comprehensive digital platform… that can model and simulate thousands of historic board games.
Model and simulate based on what?
> Nine game configurations matched the wear criteria.
So their idea was to generate candidate rulesets, have AI try to figure out rational play, then see which pieces would be moved most often and match that to the forensic evidence?
This seems impossible. If you started with chess pieces and a chessboard or even a go board and go pieces, it is absolutely impossible to reconstruct the game as they’re currently played.
Whichis is fair and good, bc nobody would expect you to rebuild the actual rules of chess from the board alone.
However it would still be useful if archeologists used the board to figure out some games similar to checkers, or go; or if they also have the pieces they could guess it was a combat game like Shogi. Any of those would give you insight about the kinds of leisure that people may get from that board.
The thing missing here is that nearly all of the games that use these nodes + stones are a variant on Nine Men's Morris (or Twelve Men's Morris, or etc). Discovering the ruleset for a particular variant is kind of like discovering a new version of "Uno" based on uncovering new house rules.
Even if the researchers did not uncover the exact right set of rules, it would probably not be dissimilar from an actual variant that could have been played somewhere in Rome.
This reminds me of bear games: https://en.wikipedia.org/wiki/Bear_games
Here's a tablebase analysis for a simple bear game I constructed a while back: https://emarzion.github.io/coqtbgen/
I really liked this game, but it frustrates me a little bit that there only seems to be one way to win. Or at least I've only found one. If I play the hardest AI, I can consistently win, with the right sequence, but it always ends up on the right side with the hares next to each other.
Is it like chess only way more obvious, that the first mover wins, and if so, wouldn't the Romans have figured it out?
Or am I just exploiting a weakness in the AI and a human would make better choices?
Reminds me of the mystery of the Roman dodecahedrons - https://en.wikipedia.org/wiki/Roman_dodecahedron
I was thinking of them yesterday because I noticed one for sale in the antiquities shop in Andor, which brought up all sorts of Earth/Rome/Star Wars cannon questions.
The best explanation anyone has come up with for those dodecahedrons was as tools for making jewelry from gold wire. Here's a video by an expert knitter named Amy Gaines in which she demonstrates the process: https://www.youtube.com/watch?v=lADTLozKm0I
The segment where she uses a replica dodecahedron to make a chain from aluminum wire (stiffer than gold, meaning this would have been easier with actual gold) is entirely convincing. It explains everything about the shape of the objects, both the protruding knobs on the end and also the fact that the holes are of different sizes.
There is a decade old paper that compiled over 50 possible explanations, and many more have been hypthesized since. Some of the dodecahedrons are solid, which may negate many of the theorized uses.
https://www.academia.edu/26809664/The_enigma_of_the_dodecahe...
I always like to hear about ancient board game reconstructions. Like music and religion, games are something every culture creates. Another recent example is the case of Liubu, a game from ancient China for which the rules were lost. Still, a reconstruction is being attempted by Carnegie Melon:
https://projects.etc.cmu.edu/liubo-lab/
Too assymetric IMO. I have no problems winning with hounds even on hard, but I have no chance of winning except against easy.
I guess you have to put it in context. There are old tic-tac-toe boards also sculpted in stone (https://legioxiiisl.blogspot.com/2013/02/roman-board-game-te...), which is a very bad game. Anything would be better.
Maybe they were playing Gobblet Gobblers instead which is a much more fun game on the same size board.
Ancient example of an unbalanced meta. Maybe we are missing some of the patch notes for balance changes
Maybe the concept of desiring balance is modern. Maybe they wanted to encode asymmetry in the game to mirror nature, or the Master / Slave dynamic which they likely considered to be a natural law.
Or they just switched sides and played again, which is less silly.
The Netrunner card game is very asymmetric and very modern. I guess you could say it's balanced but nothing asymmetric can be perfectly balanced.
Every casino game is asymmetric. I believe the casinos are trying to encode the Charlatan/Chump dynamic.
The symmetry is restored if you say you're playing a two game match and the winner is the one that survives longer as the hares. But the game seems simple enough that I imagine it shouldn't be too hard to play optimally.
I won second try with hares on medium, I think the trick is to:
A) make sure you get at least one hare to one of the central two spots (will always be possible) and camp it there. Keep your other hare moving around on your side.
B) the opponent will have to bring two hounds over to catch your other hare. To do this they will have to create a gap between the first and second hound in the middle spot above/below your camped hare. As soon as this happens, move your camped hare upwards.
It's pretty easy after that TBH. The AI has usual burnt a load of moves by that point and you have so many options from that position
I beat both on hard fairly easily? Came back to comments expecting remarks on the shape of the board allowing for fairly easy "I just move back an forth as a hare" and "I just wait for you to move wrong as a hound"
As hares, if you rush one of the middle nodes, the hound machine could capture, but it never figured out moving two hounds all the way over to do so in my play. As hounds, it's basically the same, you just make sure none of the hares ever get "behind" you because then it's a lot harder to contain them.
Since the rules are entirely reconstructed, I have to say I'm a bit skeptical of a game where a central point is keeping a count of moves up to 150. It seems too unpractical for a casual game.
Are there other known ancient games that work like this?
It should be "novus ludus" not "nova ludus". (now go and write 100 times...)
Glad I'm not the only one
i think i broke the bot! i won on hard w the hares.
figured out that my hares needed to stay together in either the top or bottom center and left corners of the big square.
at one point the bot started wasting moves so i repeated mine until it ran out! not sure if I found the deterministic win of the game or it was a bug in the AI.
The AI could have constructed any number of rules seeing as how it's literally a "best guess" and it chose rules that aren't very fun. Congratulations, I guess.
Edit: Or, put a different way: Sometimes the rules to games are lost for a reason; they're not very good.
Somewhat similar to Tafl games: https://en.wikipedia.org/wiki/Tafl_games
Similar in that it is asymmetrical, very old, and we don't quite know what the rules were (though with tafl we have slightly better historical evidence, including an account from Carl Linnaeus in 1732, which has allowed us to produce a few educated guesses). In fact tafl is sometimes speculated to derive from the Roman game of ludus latrunculorum - I'm not sure if that is the same game described here.
This game is included in Nintendo's Clubhouse Games: 51 Worldwide Classics on Switch (known in Europe as 51 Worldwide Games)
I was curious to learn more about how they discovered and reconstructed the rules. They explain that at: https://ludus-coriovalli.web.app/about
Neat, but this game seems too simple. I think AI has missed a core gameplay mechanic.
There needs to be something to stop a deadlock. Like an element of luck for the hares?
Think of this as a very old game, with a board simple enough you can trace it on the ground and play with a few stones, with rules simple enough you can explain/understand in 30 seconds. Like tic-tac-toe, mancala, nim, chopsticks... simplicity and portability wins over perfect rules or balance.
Another ancient board game to be decoded is https://en.wikipedia.org/wiki/Royal_Game_of_Ur. This is from 2600BC so is older to Christ than Christ is to us.
The videos on the game (and all his other videos) with Irving Finkel, a curator at the British museum, are spellbinding. He has the looks, manners and enthusiasm of an eccentric museum curator from central casting!
You mean Irving Finkel: https://en.wikipedia.org/wiki/Irving_Finkel
> A partial description in cuneiform of the rules of the Game of Ur as played in the second century BC has been preserved on a Babylonian clay tablet written by the scribe Itti-Marduk-balāṭu.
This is much more palatable and cool since they’re not just randomly guessing what the game can be like this article is
It's far from random. Reconstructing it with game playing AIs is a bit excessive maybe, but blocking games (hare games) are a style of game which are very widely historically attested.
They are not randomly guessing, they are doing something more sofisticated: they are correlating gameplay with the evidence of board wear.
They are creating a probability distribution function where each point is a different gameplay ruleset and then we are free to pick the gameplay ruleset with the highest probability, with the caveat that there are other similar gameplays, with slightly different rules, and almost the same probability. And there are other gameplay rules with much lower probability because they don't match the wear of the board.
A random guess would have a flat distribution where any gameplay has the same probability.
There's an online version to play that I find surprisingly addictive:
https://royalur.net/
Thank you. I had never played it, surprisingly fun. Much like Ludo but actually better.
The rules seem like a rather boring pre-decessor to the mills game/nine men's morris. I guess this is what I would draw inspiration from though, to make the game more interesting, because the board is actually kind of neat.
I’ve always been very interested in ancient Rome. I’m so happy that today, for the first time in my life, I was able to get a glimpse of it.
Nice game, had fun playing it with my friend
Playing a lost 2000 year old game is awesome.
Now I want a playable version of this game in tabletop simulator :)
Cool however no idea how to play this!