For an audience familiar with OO software engineering, the model of these diagrams could be scenarios of object instance associations, and there's standard diagrams for that. Then you could pair that with a class static object model visual that defines the associations (such as a certain association means information flow, with certain directionality, and certain roles and cardinality, and perhaps the association has attributes).
What a coincidence! I've been playing one-night werewolf with a set of friends and I was theorizing about a way to solve this problem, or at least to create a framework in which you can deduce under which "worlds" or configuration of roles some claims might hold water etc. I'm definitely going to scrutinize this article. Thanks for sharing!
edit: the article goes in particular detail about the graph structure in their own game. I'm currently looking at whether this type of analysis could be used for one-night werewolf. I was hoping for a more general and a bit more extensive post, but still. Good share.
For one night, I feel like a probability distribution function is a better idea. Something that parametrises all the revealed information and collapses into the perfect game game state if enough parameters are provided.
Without enough parameters, the PDF returns Probability(Player, Role).
I vastly prefer social deduction games that don't have a way for any player to obtain a 100% certain solution (ex: Avalon w/ Mordred). Watching your assumptions unfold over a game and having to rethink them gives a lot of room to make bold or subtle deceptive plays.
Hi, op here. The game I linked doesn't give 100 percent certainty, just a chance at asking the right questions. E.g. If you see player x talk to y, you might conclude they are collaborating. But then you ask x about it and she says after detailed discussion she worked out y was on the other team.
Meeting places are also rather subject to interception.
6 comments
[ 2.8 ms ] story [ 26.5 ms ] threadIf I had enough friends to try this out, I definitely would.
edit: the article goes in particular detail about the graph structure in their own game. I'm currently looking at whether this type of analysis could be used for one-night werewolf. I was hoping for a more general and a bit more extensive post, but still. Good share.
Without enough parameters, the PDF returns Probability(Player, Role).
Meeting places are also rather subject to interception.