To what extent do optimal strategies exist?

[cuban smoker]

With completely rational players and an known game setup, doesn't there have to be an optimal strategy for each player? It may not be obvious, as in the prisoner's dilemna, but there are just a finite number of base cases to work from (the 3 player examples in the doc thread) and even those there are more base cases than cam originally suggested, going backwards would be "simple" recursion. I agree with mathcam, I think you can O() the problem.

Now, as to your personality question, I don't think anyone here has tried to work in any kind of logic into their simulations past listening to a cop. I haven't heard of any attemps of tracking lynch votes and attempting to use history to weight the probability of anyone voting for anyone else. We all assume that lynches are random (except for some simulations which reduce the chance of lynching mafia by setting mafia to novote other mafia), and that the people are just placeholders in our model.

Extending this to the AI question, you need to analyze games to teach your AI players to play. Just like programming a computer to play chess or poker, I think you need to stick thousands of games into your database to analyze the effect of things like premature claims, refusal to claim, bandwagons, etc. on each player and the eventual outcome of the game. Then your computer will start to understand what are good plays. Next comes trying to tag each of the players in the game with personality traits, so you cansee how that weighs in. A VERY hefty enterprise if you ask me. You'd have to go through our archived games and classify each major game event as something like "claim", etc. and we don't have a lot o plain vanilla games. But if you don't use history, you're not simulating real life mafia. It would be like creating a human robot without analyzing human behaviour. It just couldn't be lifelike.

[cuban smoker]

But in your situations jeep, you can quantify the risk/reward of each reaction, and thus determine the exact correct play. That makes it possible for the Mafia considering a Lepton to determine exactly when it is to his greatest benefit to initiate the ruse. Unlike in ID's argument, which like cam said, can be completely ignored. The optimal play should have nothing to do with innanities, since every players' optimal speaking strategy will be to say "I'm townie" until it becomes optimal to claim cop.

[cuban smoker]

Fair enough. I agree completely.