[SETUP] Hollywood Mafia

[mhsmith0]

D1 stats are close to random lynching, but subsequent days are very much not

Lynched 7:2 7:1 6:2 6:1 5:2 5:1 4:2 4:1 3:2 3:1 2:1
Town 317 14 41 38 171 32 25 57 100 33 62
Scum 78 11 28 26 84 22 23 47 61 24 47
No Lynch 6 0 4 0 5 1 22 1 3 10 2
Total 401 25 73 64 260 55 70 105 164 67 111
Town 79.1% 56.0% 56.2% 59.4% 65.8% 58.2% 35.7% 54.3% 61.0% 49.3% 55.9%
Scum 19.5% 44.0% 38.4% 40.6% 32.3% 40.0% 32.9% 44.8% 37.2% 35.8% 42.3%
No Lynch 1.5% 0.0% 5.5% 0.0% 1.9% 1.8% 31.4% 1.0% 1.8% 14.9% 1.8%


Towns are a bit worse than average in 3:2 LYLO, but that shouldn't be too surprising since you're down to the subset of games where town is getting rolled anyway. In basically every other one they're above average on lynch accuracy (note that 4:2 town's accuracy is about 50% once you remove the no lynches). Town is >rand in 3:1 (~40 excluding no lynches instead of 25%), 2:1 (42% instead of 33%, though lolol that there's ever a no lynch in 2:1 LYLO), etc.

Day 1 lynch accuracy is generally difficult and not especially strong as a sign of competence. Where I question later days, though, is how much town's accuracy rate is fluffed up by the presence of PRs; I kinda suspect that things get close to rand once you back out influence of PRs, though obviously that's really hard to do.

[Toto]

You have to consider presence the PR power comes from the info from night actions, and also the fact they are a named townie. So a 3:2 with a named PR, should actually have higher chance of winning than 3:2 vanilla, even if the PR has no info.

Like I said, 7:2 odds on D1 is bad given you have PRs that should never be lynched on D1, it should be closer to 6:2.

[Toto]

According to a mix-max strategy if both team play optimally then EV is 27%, if my code is right.

@OP If you make the celebs 3:2 and 7:1 for vt:goons then EV goes up to 44.6%

[LlamaFluff]

According to a mix-max strategy if both team play optimally then EV is 27%, if my code is right.

@OP If you make the celebs 3:2 and 7:1 for vt:goons then EV goes up to 44.6%

I think the play if you want to make a setup like this is variable numbers in each group. Maybe something like:

1x Town Cop (Guilty on Goon, Innocent on Mafia Celeb)
1x Paparazzi (Guilty of Mafia Celeb, Innocent on Goon)
2-3x Town Celeb
5-6x Vanilla Town
1-2x Mafia Goon
1-2x Mafia Celeb

Otherwise the one correct lynch for town D1 basically ends the game as you are dealing with a large number of cleared players.

[callforjudgement]

Win-lose records suggest towns do worse than random in general.

Is that actually true? Like pure random lynching gives atrocious town win records; I guess the question is how towns do compared to random lynching given the power assigned to them, which I don't know the actual answer for.

My experience is that town's performance relative to EV depends on the nature of the scum's nightkill. If scum can kill whoever they like without repercussions (other than the small disadvantage they get from potentially losing the kill-WIFOM battle), towns perform below EV (by a noticable, but not huge, margin). If scum have no nightkill at all, towns perform

well

above EV (EVs can be in the 20-30 range and nonetheless manage to produce balanced setups).

In this setup, scum don't quite have a free choice of who to kill, because they may need to take out a celebrity or a non-celebrity for theory reasons, but they still have a lot of freedom. As such, I'd expect town to play at EV or slightly below in practice.

[Umlaut]

Necro-bump.

Did this ever get run? I'd like to play it, or run it if that's not an option.

[mith]

My suspicion is games like this (I call them "Census" or "Neighborhood" games, since the population of the game is split between two neighborhoods with known Mafia/Town counts) should just start with massclaim.

The strategy

at that point

is simple (in the original suggestion, with no abilities and fixed counts; might be different for variable counts):

1. Town should lynch from whichever group has a higher ratio of Mafia.
2. Mafia should kill from whichever group has a lower ratio of Mafia.

I ran numbers on these at some point but I don't know what I did with them. Something to add to my scripts.

Mostly it would play like a Vanilla game, just with a restricted pool of lynch targets at any given time (and switching between the two groups as Mafia are lynched).

[Umlaut]

mith, what is there to claim in this setup? Everyone already knows who is a celebrity, and no one is going to claim they're scum.

[mith]

Sorry, I was talking about a more general version of the setup (where the number of town celebrities is known but not confirmed). Misread the OP (but it doesn't matter - they work out to be the same game).

(IMO, the general version is very slightly more interesting, in that Mafia have a choice of who is pretending to be a celeb - but less thematic for Hollywood, obviously.)

The important thing (for EV calculation) is getting the lynch/kill strategy right. (From that point of view, the concern about a lucky D1 lynch is somewhat unfounded - it isn't optimal for town to lynch from the Celebs, so they should have the lucky lynch of hitting the Mafia Celeb D1.)