Mafia: Theory

Use this topic to discuss theory. This way it'll make it easier to identify the Tips and Tricks in that thread, and we can hopefully avoid confusing any newbies.

Wagoning them early in the day ends the day prematurely and deprives the Town of the opportunity to discuss further.
What's to discuss? Killing the scum is town's primary objective, and helps prevent getting into MyLo/LyLo (or might get you out of it). As such, taking a direct step towards it is extremely difficult to justify not doing. Chasing after something else is just dumbass play that indicates that the people proposing doing anything else are scumaligned or idiots.
Lynch the idiots would be another good tip, I suppose. Stupid play is a liability, nearly as bad as lying.

Hey, I resent this piece of advice.
An untested hypothesis, if I were mafia on mafia 3, my coconspirators would have advised me against stupid play.
Maybe the idiots have a lower chance of being mafia?
Of course, if communication will be limited on mafia IV, we can have mafia players xaading around.


Well, the shoe fits, I can't deny, but I'll improve my play on the next one

What's to discuss? Killing the scum is town's primary objective, and helps prevent getting into MyLo/LyLo (or might get you out of it).
Daytime is a resource too. There's a reason why scum generally wants to end the day early, even to the point of selfhammering  because the sooner the day ends, the less opportunities the rest of the scum team has to slip up or be pressed for explanations.
Let's say you have two obvscum players, and you wagon them at the start of day 1 and 2. Then day 3 comes, and you're completely in the dark, because there was zero discussion on the previous days. OTOH, if you used the time, you likely have some idea on people's behavior.

Let's say you have two obvscum players, and you wagon them at the start of day 1 and 2. Then day 3 comes, and you're completely in the dark, because there was zero discussion on the previous days. OTOH, if you used the time, you likely have some idea on people's behavior.
I think that's a pretty weak argument. Or rather yes, you can discuss as much as you want, but still hammer the scum. Nolynching because you were hoping someone might say something useful (which is a gamble at best) is very poor play as it gifts the other side with another night, and lynching at random (or even a known town!) when there's a strong suspect about is unlikely to go well.
You're trading definite actions for gambles and unlikely chances. Those odds are poor.

Let's say you have two obvscum players, and you wagon them at the start of day 1 and 2. Then day 3 comes, and you're completely in the dark
Counter argument: let's say you have two obvscum players, so, sure there's more out there, you press people and hem and haw and lynch someone you're not sure about because time is running out on day 1, and lynch based on some investigation results (unbeknownst to you, from scum 3) on day two. Now you've given the three scum time to plan, when it would have been one scum on his own, and you've lost four Town.
That's much more how our Town plays

Now you've given the three scum time to plan, when it would have been one scum on his own, and you've lost four Town.
Yeah, but you've also given the town two obvious lynch targets which they can go after despite the losses.
OTOH  you hammer the two obvious scum, and by day 3 the rest of the Mafia has taken out the experienced players that could push for the lynch, leaving you with lurkers. And you're either in nolynch hell, lynching randomly, or lynching whoever the scum wants you to lynch. You can't even analyze the bandwagons from D1 and D2, since if they're obvscum, it doesn't matter for the Mafia if they jump on it or not.

Yeah, but you've also given the town two obvious lynch targets which they can go after despite the losses.
So, do you plan to start lynching scum before or after they win?

Let's say you have two obvscum players, and you wagon them at the start of day 1 and 2. Then day 3 comes, and you're completely in the dark, because there was zero discussion on the previous days.
FFS, @Weng, can you have the next random encounter be a neon sign saying "Yami and DoctorJones are scum!"? Because I think it's the only thing that could balance the game to at least give the town some fighting chance.
I mean, DoctorJones has all but hung that sign over himself yesterday, and they just ignore it, and Yami pretty much already has that sign over herself since that cop noite, and you're convinced there's only two scum  which is wrong, but you'll find it out two days later when the scum narrows the pool of suspects down by two.
For now, why the fuck aren't you lynching the two? Off Yami, off DoctorJones with an additional clause that whoever refuses to vote is scum since he's loved, bam, game won (or so you think).
!vote for Maciejasjmj

Hey, I'm learning too.
Besides, back then they were thinking there are only two scum left  so duh, if you have two obvscum people and you think there are two left, then yes there's no point in looking for scum anymore, is there?

If you lynch two scum, on day 3 you may have no information, but that's just as bad as you would normally be on day 1, except with two less scum. At that point you can do all the things you wanted to do day 1 to pressure people into voting, but with two less scum.

but with two less scum.
That doesn't matter as much as having two townies less. And a goodplaying scum team will make sure to off those two people who would lead the investigation later on.

Killing 50% of scum is less important than killing 25% of townies? (Numbers for a 12player game with 4 scum)
I'm honestly asking, because I suck at the statistics.

Obviously you'll lynch them eventually.
Point is, given that you know when you're at MyLo/LyLo, any bunch of idiots can lynch the obvscum pair when it's necessary. It's kind of important to use your experienced players while they're alive  and that means going after less obvious targets first.


@Maciejasjmj said:
Obviously you'll lynch them eventually.
“Obviously”
Well, yeah, it's kinda in the rules that you have to get rid of all scum before they win. If you're at LyLo/MyLo, then all bets are off.

TIL that @Maciejasjmj is weak to WIFOM arguments.

my coconspirators would have advised me against stupid play.
Town advised you against stupid play.......
So....

I'm honestly asking, because I suck at the statistics.
Statistics don't help, because on day 1, you have a better statistical outcomes by not lynching. The statistical outcome improves for lynching as more people die.


Statistics don't help, because on day 1, you have a better statistical outcomes by not lynching.
Where did you get that? No you don't. And if the town would have a better chance of overall win if they nolynch and let Mafia kill (which is occasionally true  you have a better chance of winning with an even number of players), then Mafia can simply not kill and you're back to square one.

Day 1 is different from all other days.

Well yeah, no lynching on day 1 is terrible, no lynching on any other day is awfully terrible.

Statistics don't help, because on day 1, you have a better statistical outcomes by not lynching. The statistical outcome improves for lynching as more people die.
Actually, that somewhat depends on the ratio between town and mafia, as well as the overall number of players. According to wiki.mafiascum.com^{[1]}, an odd number of players has a slight statistical edge for the mafia, while an even number of players has a slight statistical edge for the town.
As @aliceif mentioned, Day 1 is unique, and allows the opportunity for the town to shift the odds by judiciously choosing to not lynch. Under the right circumstances, choosing to not lynch on Day 1 can slightly improve the statistical odds of winning for the town.
However, it should be kept in mind that these statistics are based on purely random lynchings and mafia kills. They do not take into account actual game play that could easily shift the balance of power either way.
^{[1]} See ...and they all lived happily ever after and Numbers, Part 1.

10T 4M
Outcomes.
4 x (1 Town Dead + 1 Mafia Dead)
10 x (2 Town Dead)vs.
10 x 1 Town Dead

So, basically, lynches random means:
Lynch 
28.5% day2 = 9T 3M
 25% day3 = 8T 2M
 75% day3 = 7T 3M
71.5% day2 = 8T 4M
 33.3% day3 = 7T 3M
 66.7% day3 = 6T 4MNo Lynch 
100% day2 = 9T 4M
 30.7% day3 = 8T 3M
 69.3% day3 = 7T 4MSo, if you lynch on day 1, you have a 23.8 (miss + hit) + 28.5 (hit + ) = 52.3% chance of getting at least one mafia in the first two days. Without accounting for any information gained from the day 1 wagon or investigations in night 1 or so forth and so on.
If you don't lynch on day 1, you have a 30.7% chance of getting at least one mafia in the first two days.
So lynching on day 1 improves your chance of being 1 mafia ahead within 2 days by ~21.5% (percentage points)

But 1 Town Dead != 1 Mafia Dead.
10T 4M:
Outcomes:
4 x (10% town power loss, 25% mafia power loss)
10 x (20% town power loss)vs
10 x (10% town power loss)
You have a 28% chance of adding a 25% mafia power loss on top of your forsure 10% town power loss, at the cost of another potential 10% power loss.

chance of getting at least one mafia in the first two days. Without accounting for any information gained from the day 1 wagon or investigations in night 1 or so forth and so on.
At the risk of doubling your losses.
But 1 Town Dead != 1 Mafia Dead.
This is more convincing.
But the problem comes in that there are so many variables that it's impossible to tell.
What if you kill your blocker? That's more than 10% loss.

What if you kill your blocker? That's more than 10% loss.
What if you kill one of them that's in a lover's pair with another? That's a 50% gain.
In a closed setup, you can't ever tell.

If it's an open setup, you can work it all the way down, which the numbers link @abarker provided has done for several common open setups at MS.
Given that we're playing closed setup, it's a risk you must take.
Because even if you assume that the 10/4 town nottown setup is a given, you have no idea whether we're playing 10T + 4M, or 10T + 2M + 2M' (and the possibility of 2 town losses a night), or 10T + 3M + Arsonist (and if you stall, suddenly the Arsonist is getting free days to prime and, boom, you get @Macie'd).

@vote for @maciejasjmj this shit doesn't make sense

Game does not exist
@PleegWat tried to vote for maciejasjmj in post #52.!for @maciejasjmj
<! SockBot/2.11.9 (Cheery Chiffon; owner:yamikuronue; user:vote) 20160121T17:35:54.067Z >

OMG we should totally have a meta Game going on in the theory thread!


So, basically, lynches random means:
Uh oh! Statistics! We're dangerously close to getting the tables…
The first play should be the pick of the dice gods in the absence of other evidence. But as evidence — whether direct or by close observation of people's statements and behaviours — accrues rapidly, the whole “look at the statistics of a random model” approach rapidly becomes just so much horseshit. You're not playing an ensemble of random setups or against a group of memoryless bots; you're playing one setup and against people.
Is it time to revive the Monty Hall discussion?

Anyone wanna point to information about how the game's played, and what things like wagoning and mylo/lylo mean?

http://wiki.mafiascum.net/index.php?title=Main_Page should have them

Oh shiznit, this is everywhere now.
NHS is watching you metascum.

Look it's obvious there are metascum here, planning their future scumming, by confusing potential townie players.

NHS is watching you metascum.
I don't think so. I think they're having a panic over the doctors' strike. [spoiler]You did mean National Health Service, yes?[/spoiler]<! and fuck you discourse >

Lynch the idiots would be another good tip, I suppose. Stupid play is a liability, nearly as bad as lying.
But how do you know if the play is really stupid?

But how do you know if the play is really stupid?
You don't. But if they're playing an idiot on TV, that'll do.

So, look for Tony Robinson?

So, look for Tony Robinson?
But his plans aren't stupid. They are cunning! There is a difference...

The criteria was "playing an idiot on TV", you ask for miracles, I give you a character from the BBC.


Yippie ki yay, Mr B?

10T 4M
Outcomes.
A bit too much. Let's go for something smaller  a 52 setup.
Green edge means scumlynch, red edge means town lynch, blue edge means nolynch.
Town L means probability of town lynch, Win L means probability of winning after a random lynch (and assuming optimal play from then on), Win NL means probability of winning after not lynching. Also assumes the Mafia kills each night.In short  an odd total number of players generally means that even a random lynch is better than not lynching at all. An even total number of players slightly favors a nolynch, but it assumes the Mafia will kill someone overnight to bring the total to an odd number.
And in either case, random lynching nearly universally gives odds stacked against town.