Charisma Model of EV Calculation

[popsofctown]

the important thing is how should setups be balanced right

with random lynching gen consensus was ev ~40% because we assume town can be big brain to help out wrs

do we start balancing @50% using this model or ... ?

You shouldn't balance to 50%, imo, but that aside, you should balance to random EV or greater.

This metric is just describing a phenomenon that is somewhat prevalent and the OP demonstrates that it's wildly suboptimal.

It's trivially possible to perform at random EV levels, so any balancing exercise should treat that as the lower bound even if poorer strategies exist.

This model is better for balancing even though it's based on a suboptimal strategy because it mimics the way people actually play. As you add mechanics to this it's going to model real gameplay better than random lynches. For instance, in a theoretical setup where players could grant bulletproof and tracker to a player day 1, the random lynching model sometimes has the town change their mind and lynch that player and drop EV, while this model acknowledges that it's more likely than not that player continues to get townread and that tracks continue to get fired and increase town EV. You can change the random lynching model to say "uh town never lynches the guy that got bulletproof", but then at that point you're partially adopting this model because you know it's better.

Mostly-static-sorting

is

the optimal strategy in real games because scumreading scum is real, and purely random lynching doesn't align with a reality where scumreading scum is real.

You should definitely ignore percentages though.

[zoraster]

this is pretty neat, mathdino.

[Psyche]

it's got me all excited about the game again
really opens my mind about what's possible

[Kagami]

the important thing is how should setups be balanced right

with random lynching gen consensus was ev ~40% because we assume town can be big brain to help out wrs

do we start balancing @50% using this model or ... ?

You shouldn't balance to 50%, imo, but that aside, you should balance to random EV or greater.

This metric is just describing a phenomenon that is somewhat prevalent and the OP demonstrates that it's wildly suboptimal.

It's trivially possible to perform at random EV levels, so any balancing exercise should treat that as the lower bound even if poorer strategies exist.

This model is better for balancing even though it's based on a suboptimal strategy because it mimics the way people actually play. As you add mechanics to this it's going to model real gameplay better than random lynches. For instance, in a theoretical setup where players could grant bulletproof and tracker to a player day 1, the random lynching model sometimes has the town change their mind and lynch that player and drop EV, while this model acknowledges that it's more likely than not that player continues to get townread and that tracks continue to get fired and increase town EV. You can change the random lynching model to say "uh town never lynches the guy that got bulletproof", but then at that point you're partially adopting this model because you know it's better.

Mostly-static-sorting

is

the optimal strategy in real games because scumreading scum is real, and purely random lynching doesn't align with a reality where scumreading scum is real.

You should definitely ignore percentages though.

Random EV usually implies a sort of sensible policy beyond simply behaving random in games that involve roles, such as making the assumption that your BP tracker is town in your example (and accepting the loss if wrong).

Static sorting is not optimal because scum can selectively kill players who have correct reads, and so static reads conditioned on being alive end up being quite bad. Almost everyone has reads that are very close to chance (and almost everyone will cherry pick their own performance to argue otherwise).

[popsofctown]

Shooting players with correct reads can be counterplayed by deadsheeping.

In practice, people don't deadsheep as much as they should, but also in practice, people have mostly static reads, so like.. this is a good model.

[RadiantCowbells]

People have a strange tendency, no matter how poor their read record, to consider their own read on a player to be the single most important tool to determine whether to vote them or not

[OkaPoka]

More fun to trust your own reads

[Bitmap]

More fun to trust your own reads

ok YukzYuk

[Ramcius]

People have a strange tendency, no matter how poor their read record, to consider their own read on a player to be the single most important tool to determine whether to vote them or not

Admitting that your reads sucks means admitting that you're bad at mafia, no one want that

[callforjudgement]

People have a strange tendency, no matter how poor their read record, to consider their own read on a player to be the single most important tool to determine whether to vote them or not

One of the biggest improvements in my townplay came from getting better at figuring out how confident I should be in my own reads (and being honest about this to the rest of the players). There's a spectrum between hyperpushing your own reads very hard, and disregarding your own reads entirely in favour of listening to other players and asking them for help. Good play consists of working out where on that spectrum you should be, and it's often different from game to game, and from day to day within a game; both extremes are correct sometimes, and recognising when is likely a very important factor in winning as town.

[Dannflor]

what if you can have sucky reads but also be good at mafia

[Nancy Drew 39]

Can someone please tell me what “EV” means?

Thanx.

[Dannflor]

https://wiki.mafiascum.net/index.php?title=EV

[Nancy Drew 39]

https://wiki.mafiascum.net/index.php?title=EV

Thank you.

[Nancy Drew 39]

My assumptions:
Gamestates result from predetermined charisma levels, where charisma is defined as "ability to avoid being lynched".
Town will always attempt to lynch the least charismatic (scummiest) player. Town will never no-lynch.
Scum will always attempt to kill the most charismatic (towniest) town player.
PRs will claim at L-1. If uncc'd, they become most charismatic.
Scum will fakeclaim confirmable roles at L-1 in order to out TPRs, and will be lynched if counterclaimed.
Scum are aware of the charisma list, and will counterclaim PRs if they are more charismatic and if doing so would not lose them the game.

Scum on they’re way down might also attempt this CC gambit. In my my first scumgame on MU (third anywhere), we had our least charismatic scum member, do just that, because if they were inevitably going down anyway, might as well help out the tpr, right?

ETA: They were not at L-1 att but we saw no way to save them.