something i heard radiantcowbells talk about (and maybe there is a thread where he hashed this out that i don't remember) is approaching EV by laying out an ordered sequence of the players in terms of towniness, randomly placing the mafia within that sequence, and then evaluating how the game would play out under those circumstances, where the least towny player gets executed each day. imagining a nightless game with three mafia and six town, mafia wins as long as one of them is at least the second towniest (game ends @ final 3), OR if two of them are the third and fourth towniest (game ends @ final 5); OR if three of them are in the top 6 towniest (game ends at final 7).
for an example with nightkills, consider a similar-EV setup of 2 mafia versus 11 town. in addition to the scummiest player getting executed each day, the towniest town player gets executed every night. thus, mafia now wins if they have one player in the towniest 6 (!) (endgame in final 3 after the top four players are shot and the bottom four players are executed) or the 7th and 8th places (in final 5 after 3 are shot and 3 are executed).
suddenly this setup looks kind of brutal for town. now, obviously there's various ways you could argue that this model is less descriptive, the primary one being that it assumes perfectly effective nightkilling (which is hard to do all four nights). but it spells out the way that nightkills are meaningfully harmful to town's EV.
to do:
- ev calculation for dragoneater's setup
- make a thingy in R that spells out different setups' sensitivity to good reads (i.e., how do things change if wolves are distributed lower in the towniness pile)