A ludeme is an abstract fundamental unit of play in a game, such as the L-shaped movement of a knight in chess, as opposed to an "instrument of play" which forms part of the equipment with which a game is played, such as the knight piece itself. It is used to describe games between game scholars or to AI, and is a part of General Game Playing AI systems that are used to analyze ancient games.
Ludemes have three main properties that remain consistent between all users of the term:
Additionally, the term can be differentiated based on if the ludeme can be broken down:
By this definition, entire games can be categorized as ludemes because they are discrete, transferrable, and contrastive elements of play that are made up of sub-ludemes. The use of Compoud Ludemes, however, is not universally agreed upon.
The term was originally coined by French game writer Pierre Berloquin. Alain Borvo, one of the first to use the term in writing, defines it as a 'type rule' such as the method of trick-taking in a card game or the leap capture in a board game.
Ludeme as a term is considered similar to the term "meme" coined by Richard Dawkins in The Selfish Gene, and as such it may also be referred to as a "game meme" because both have the property of transmitting throughout human culture. While the two terms bear similarities, it is unclear if the term ludeme was derived from Dawkins's definition of a meme.
Ludemes are used as the foundation of the Ludii General Game System which is programmed in Java. Using Ludii's Game Description Language, a game can be written out using ludemes in a way that a computer can interpret to make the game playable by humans or AI. Ludii is built using a class grammar approach, where each ludeme described corresponds to a Java class of the same name.
The Ludii General Game System is primarily used for developing General Game Playing AI that can play any game, as opposed to AI designed to play a specific game. Ludii is generally considered easier to write and read compared to it's counterpart General Game Systems such as the Stanford Game Description Language (GDL) and Regular Boardgames (RBG).
The Ludii system uses a specific syntax to write out games using ludemes. Each Ludii game description contains information about the game's Players, Equipment, and Rules, in roughly the following structure.
For Example, Tic-Tac-Toe described in Ludii syntax would look as follows:
(game "Tic Tac Toe"
( players 2 )
( equipment {
( Board ( square 3 ) )
( piece "Disc" P1 ) ( piece "Cross" P2 )
})
( rules
( play ( move Add ( to ( sites Empty ))))
(end if ( is line 3 ) ( result Mover Win ))
)
)
This description is broken down as follows, and states that the game:
The Ludii program also tracks the "game state," which changes based on the "Play" actions that are taken by human or AI players.
The Digital Ludeme Project is a research project launched by the University of Maastricht in 2018. It's goal is to reconstruct ancient games using the Ludii General Game System in combination with AI systems. The rules of ancient games were commonly transferred orally and not written down, so the project aims to have AI automatically play games using the Ludii software to identify likely rulesets based on existing information, such as recovered game pieces.
The project has a database of 1057 games described in the Ludii Game Description Language and contains information about a game's ruleset, equipment, geographic location it was found, and time period. AI systems using the Monte Carlo tree search algorithm then play those games and record statistics, such as the game length, number of pieces on the board, and win-loss ratio of each player. These statistics are used to infer the "quality" of a given ruleset based on the likelihood that a human would enjoy the game (For example, games with a skewed win-loss ratio or longer play time are considered lower quality). While playing, AI can also identify issues with a given ruleset, such as if it results in an infinite loop.
These statistics can be used to identify likely rulesets for a game whose full rules are unknown. Researchers can also use the database to identify games that share common rulesets within a similar geographic region, allowing them to track the transmission of ludemes and possible evolution of games throughout history.