S-FM 218: Mafia Wars III

[BananaCucho]

I should actually write an algorithm that does that. Should only be extremely difficult. But worth it to stop this madness

Ill write one specific for this game and run it a thousand times.

[Firebringer]

Super-stable? This game is inherently going to be unbalanced.

You have scum being able to pick their powers - no way can this be accounted for balance-wise.

Luciano has the advantage because they can tailor their roles to the strategies that they, as players, will take towards the chat.

FBI is weaker because they can't actively choose which abilities they will end up with and thus can't plan ahead as well.

This should be taken into account.

I mean scum picking powers could be taken advantage of if day 0 the scum picked and then host made the setup.
Would be pretty easy to change that part of setup.
GF and Consig are stuck in so you just base rest of roles around what the Luciano have.

Only problem is that Luciano know that, but then hey theres still some room for mod WIFOM.
In this you kind of no idea if the roles you pick will really help you more because you don't know what town has.

[Calix]

GF/ Consig have been removed from the setup. I'm not sure how well-balanced the updated setup is so I would give that a check.

[MattZed]

60% isn't an algorithmically generated figure. In fact, I'm much opposed to algorithmic figures for Mafia chances because they're incredibly sensitive to assumptions about town's chance of lynching correctly, which can vary wildly. Rather, I get my figure from thinking of similar two-scum games that I might consider balanced and changing percentages relative to how much more balanced or unbalanced I think this setup is than that one. And I find this a fair bit more townsided, as reflected in the 60%.

One way to get at that is to ask "how much better or worse could this setup have gone for town?" My general conclusion is that it could have gone much worse; although there are things that could have helped town even more, (the existence of a cross-kill, or if the Kidnappers had chosen to jail scum instead N1 and executed them then) most alternative versions of this game have town doing worse, often considerably worse. (kidnappers jail and execute a town who didn't claim GF/consig, kidnappers don't get to act, Lucianos choose better kills or had chosen better roles for this town balance, any amount of mislynches had occurred)

[BananaCucho]

That's a lot of words for:

"I pulled that number out of my ass"

[MattZed]

That's a lot of words for:

"I pulled that number out of my ass"

Sure, but I'm using it to quantitatively express my feelings about the setup. There's a difference between a game where town will win 95% of the time and one where it will win 50% of the time, but both are townsided. I'm not introducing numbers as a fact, but rather so we can gauge where we think the balance is.

[Firebringer]

60% isn't an algorithmically generated figure. In fact, I'm much opposed to algorithmic figures for Mafia chances because they're incredibly sensitive to assumptions about town's chance of lynching correctly, which can vary wildly. Rather, I get my figure from thinking of similar two-scum games that I might consider balanced and changing percentages relative to how much more balanced or unbalanced I think this setup is than that one. And I find this a fair bit more townsided, as reflected in the 60%.

One way to get at that is to ask "how much better or worse could this setup have gone for town?" My general conclusion is that it could have gone much worse; although there are things that could have helped town even more, (the existence of a cross-kill, or if the Kidnappers had chosen to jail scum instead N1 and executed them then) most alternative versions of this game have town doing worse, often considerably worse. (kidnappers jail and execute a town who didn't claim GF/consig, kidnappers don't get to act, Lucianos choose better kills or had chosen better roles for this town balance, any amount of mislynches had occurred)

You do a worst case analysis of your setup from all sides.

In this game, the worst case scenario here for scum:
N0 FBI fails to recruit
D1 No Lynch so Kidnappers jail
N1 Both kidnappers kill a scum
D2 Last scum lynched.

In this game worst case scenario for town:
N0 Kidnapper recruited by FBI
D1 No lynch
N1 Kidnapper kills town both kidnappers, Both scum kill [4 deaths]
D2 Mislynch [1 death that is 8 left]
N2 2 deaths again [6 left]
D3 Mislynch [5 left]
N3 Last town killed and scum or possibly tie with scum.

So basically even in worst case for town, they make it to night 3 and the scum draw.

[Firebringer]

You do a worst case analysis of your setup from all sides.

In this game, the worst case scenario here for scum:
N0 FBI fails to recruit
D1 No Lynch so Kidnappers jail
N1 Both kidnappers kill a scum
D2 Last scum lynched.

In this game worst case scenario for town:
N0 Kidnapper recruited by FBI
D1 No lynch
N1 Kidnapper kills town both kidnappers, Both scum kill [4 deaths]
D2 Mislynch [1 death that is 8 left]
N2 2 deaths again [6 left]
D3 Mislynch [5 left]
N3 Last town killed and scum or possibly tie with scum.

So basically even in worst case for town, they make it to night 3 and the scum draw.

This only takes into account the kidnapper added KPN, not in any other roles for the record.
Scum can lose by day 2.
Town can only lose by Day 4 if scum never cross kill and alot of fucked up shit happens.

[BananaCucho]

With an algorithm what you do is plug in the roles you picked

Then everything else is random. Lynch or no? This is the hardest part. I say take all recent games that exist on site and figure out percentage of days that ended in lynch vs no lynch and scale it that way and make each day random (special cases taken into account for LYLO and such). Yeah each game is going to vary but if you run it enough (a thousand times should do it) you should be able to calculate a fairly average number of where your setup is likely to lie.

Of course if it's setup specific you can factor in logic that says "if GF reveals he doesn't get lynched". You can get a pretty precise number this way.

[SuperJack]

God this argument is worthless. Why don't you both just have a cup of shut the fuck up..

[MattZed]

This only takes into account the kidnapper added KPN, not in any other roles for the record.
Scum can lose by day 2.
Town can only lose by Day 4 if scum never cross kill and alot of fucked up shit happens.

In the worst case scenario you outlined, I would consider town to have lost by N2. The situation you put in assumes town goes into N2 in a 4v2v2 and comes out of it 2v2v2 against two factions with night kills. Town having any reasonable chance of pulling that one out? Yeah.... right.

Although I disagree in general with worst case scenario analysis because it should almost always take longer for a perfect scumteam to kill the town than for a perfect town to kill the scum by virtue of numbers, and also doesn't account for how these outcomes might change probability.

[BananaCucho]

God this argument is worthless. Why don't you both just have a cup of shut the fuck up..

=)

[SuperJack]

[MattZed]

With an algorithm what you do is plug in the roles you picked

Then everything else is random. Lynch or no? This is the hardest part. I say take all recent games that exist on site and figure out percentage of days that ended in lynch vs no lynch and scale it that way and make each day random (special cases taken into account for LYLO and such). Yeah each game is going to vary but if you run it enough (a thousand times should do it) you should be able to calculate a fairly average number of where your setup is likely to lie.

Of course if it's setup specific you can factor in logic that says "if GF reveals he doesn't get lynched". You can get a pretty precise number this way.

My point is that, while I agree you can do this, I strongly disagree with making assumptions about "town tends to lynch x% better or worse than chance" because the imprecision in this figure almost single-handedly determines if the algorithm says the setup is balanced.

[Firebringer]

In the worst case scenario you outlined, I would consider town to have lost by N2. The situation you put in assumes town goes into N2 in a 4v2v2 and comes out of it 2v2v2 against two factions with night kills. Town having any reasonable chance of pulling that one out? Yeah.... right.

Although I disagree in general with worst case scenario analysis because it should almost always take longer for a perfect scumteam to kill the town than for a perfect town to kill the scum by virtue of numbers, and also doesn't account for how these outcomes might change probability.

I mean the scum factions can't win together.....
They are going to try to cross kill each other.
So no, I don't think its reasonable to say that town is out because they no longer have majority.

[SuperJack]

My point is that, while I agree you can do this, I strongly disagree with making assumptions about "town tends to lynch x% better or worse than chance" because the imprecision in this figure almost single-handedly determines if the algorithm says the setup is balanced.

[SuperJack]

I mean the scum factions can't win together.....
They are going to try to cross kill each other.
So no, I don't think its reasonable to say that town is out because they no longer have majority.

[BananaCucho]

My point is that, while I agree you can do this, I strongly disagree with making assumptions about "town tends to lynch x% better or worse than chance" because the imprecision in this figure almost single-handedly determines if the algorithm says the setup is balanced.

Except the algorithm doesn't say what chance of town lynching correctly. It assumes any of the 13 players is lynched except on special cases like may or reveal. 100% random

What I was data collecting is chance of anyone being lynched vs no lynched. We have data for that based on site meta. Thats not 50%

[SuperJack]

Except the algorithm doesn't say what chance of town lynching correctly. It assumes any of the 13 players is lynched except on special cases like may or reveal. 100% random

What I was data collecting is chance of anyone being lynched vs no lynched. We have data for that based on site meta. Thats not 50%

[MattZed]

I mean the scum factions can't win together.....
They are going to try to cross kill each other.
So no, I don't think its reasonable to say that town is out because they no longer have majority.

Town needs to, without controlling the lynch, kill 4 scum while only having two of their own. Their only way to win is to lynch one of the FBI, have the other FBI shoot a Luciano at night, have the Luciano that wasn't night killed kill the FBI, and then the town have to figure out who each other are in a 2v1 LYLO.

This will basically never happen.

[Firebringer]

Town needs to, without controlling the lynch, kill 4 scum while only having two of their own. Their only way to win is to lynch one of the FBI, have the other FBI shoot a Luciano at night, have the Luciano that wasn't night killed kill the FBI, and then the town have to figure out who each other are in a 2v1 LYLO.

This will basically never happen.

ummmm....it has happened in multiple scum team games.

[MattZed]

Except the algorithm doesn't say what chance of town lynching correctly. It assumes any of the 13 players is lynched except on special cases like may or reveal. 100% random

What I was data collecting is chance of anyone being lynched vs no lynched. We have data for that based on site meta. Thats not 50%

I disagree with an assumption that lynches basically happen at random. In fact, the data we have would suggest that lynches are actually a fair bit better than random on average. (Purely random lynches says 10 cits vs 1 SK is an incredibly balanced game, but experience suggests otherwise)

[MattZed]

ummmm....it has happened in multiple scum team games.

Quote one.

[Firebringer]

Quote one.

it would be offsite games.

[MattZed]

it would be offsite games.

I'm seriously interested to see the game where town wins a 2v2v2 where the scum factions each have a factional kill.

[Firebringer]

I'm seriously interested to see the game where town wins a 2v2v2 where the scum factions each have a factional kill.

I am too lazy to point these out ill just let someone else who should know of this being a thing:
@ika want to back up my claim here?

[BananaCucho]

I disagree with an assumption that lynches basically happen at random. In fact, the data we have would suggest that lynches are actually a fair bit better than random on average. (Purely random lynches says 10 cits vs 1 SK is an incredibly balanced game, but experience suggests otherwise)

You misunderstand the idea behind randomness

Given 1000 runs of the same game sometimes you have strong town, sometimes you have weak town

Run it 1000 times, collect the average, there is your percentage.

[SuperJack]

HEY MATTZED

I KILLED A PUPPY.

HAD YOUR SETUP BE PERFECTLY BALANCED IN EVERY SCENARIO THEN I WOULD OF NOT KILLED THE PUPPY.

SO THE PUPPYS DEATH IS YOUR FAULT

[Firebringer]

HEY MATTZED

I KILLED A PUPPY.

HAD YOUR SETUP BE PERFECTLY BALANCED IN EVERY SCENARIO THEN I WOULD OF NOT KILLED THE PUPPY.

SO THE PUPPYS DEATH IS YOUR FAULT

yeah matt, stop making him kill puppies.

[ika]

Town needs to, without controlling the lynch, kill 4 scum while only having two of their own. Their only way to win is to lynch one of the FBI, have the other FBI shoot a Luciano at night, have the Luciano that wasn't night killed kill the FBI, and then the town have to figure out who each other are in a 2v1 LYLO.

This will basically never happen.

it has happened before on offsite games. ive seen a 1v1v1 game too. know what happened? town would not vote and casues a no lynch, scums cross killed an town won.

you of all people should know that if it can happen it can happen not matter how unlikely. thats how mafia is. when you are arguing worst case scnarios you are arguing something that almost never happens so trying to apply it here is double standard

[SuperJack]

yeah matt, stop making him kill puppies.

I ALSO JUST KILLED A PUPPY. BECAUSE YOU TROLLED ME.
YOU USED NO CAPITAL "Y" AND CALLED MATTZED MATT.

I HAVE JUST KILLED ANOTHER PUPPY.

THIS IS YOUR FAULT.

[Firebringer]

I ALSO JUST KILLED A PUPPY. BECAUSE YOU TROLLED ME.
YOU USED NO CAPITAL "Y" AND CALLED MATTZED MATT.

I HAVE JUST KILLED ANOTHER PUPPY.

THIS IS YOUR FAULT.

thats okay, im not a big puppy fan anyways.

[Gyrlander]

thats okay, im not a big puppy fan anyways.

Fucking murderer.

[SuperJack]

See people, It's not my fault at all that I'm Murdering these Puppys. They are the real guilty party.

Just so you all know, I'm not Murdering these Puppys to hurt any of you in any way, knowing full well that these puppy deaths are horrifying and will ruin your night.
(They are, I checked google out for some pics to post with, but they are too bad).

[Yukitaka Oni]

Lel

[PowersThatBe]

I liked fire ringer too. That was disappointing to say the least.

[Whad]

Ika/silver and Whad really need to research what a slip is. Those last 10-15 pages before EoD were disguntigly bad plays. Hypocritical and inconsistent dont even begin to describe what went down. Would rather have a lobotomy than play with either of you again until you can self reflect and you gain some concept of critical thinking.

#rantover

Rest of game was very enjoyable. Thanks to MZ for hosting

@Iced_Monopoly
I'm sorry you feel that way. I guess ika was trolling you, but I was deliberately trying to rile you up to get you to post more loosely so I could get a better read on you. I guess it worked, since you still seem upset about it. Until our interactions, I thought there was way too little content from your slot and your calculated posting style made it that much harder to read.

So I'm sorry you found that part of the game unenjoyable. I felt quite the opposite. Some of your archon's posts (I can't remember who posted them) had me audibly laughing.

[Whad]

GAIZ, WE DID IT, TOWN WON THE G-



Lol, what the fuck Firebringer.



This game was won by the Town fairly.
The scenario in which two kidnappers are unbalanced, is if we think that the Luciano never thought of using them, as well as them conviently getting a No Lynch and both jailing the same target who purposely initiates a circle to confirm ourselves.

Even without Firebringer gamethrowing, I would lead the Town into a victory lol, even if Yuki decided to kill me and Firebringer cleaned me - the two jailors + consig would just reveal themselves to confirm me and the clean was just plain useless anyway. I'm the most confirmed thing ever, so are the two jailors - the only one that wouldn't know anything is the Consig because they were not part of our secret discussion, and that hole would be plugged assuming that Whad took my message into establishing codes.

The entire outcome of this game would change if I wasn't double jailed, if there were two seperate jail targets then this might have been an obscured fact and only one jailor would be revealed while the other is incognito - slowing down strategy significantly.

RLVG, Elephant and YZB shared MvP lol.

Firebringer had absolutely no reason to troll himself into a loss, he assumed the game was unbalanced - but it wasn't that severely unbalanced, it's a very situational balance in a Mystery Setup, it was all pure luck which could have had the tide turned with good clean targets. Trudeau being cleaned is THE worst move they could've done that night and in fact the Rule Of Dumb would've lead Trudeau into being anti-town with benefit of Scum in the longterm.

The whole FA's emotional situation was also game altering, but of respect I won't discuss that part further.



Fun fact : I sendt in my revisioned LW and reveal, but because of database errors it got rollbacked and was never sendt.



Setup balance :
Make FBI immune to night kills for the first few nights (buff).
Make a person only be targeted by a single jailing (nerf), and it being prioritized by Luciano (buff).

I agree with this post in that a huge reason why town was so strong was because of the extenuating circumstances. If Elephant hadn't been kidnapper, and able to confirm his role by jailing, we might have lynched him anyway. If Elephant and yzb25 didn't jail RLVG then they wouldn't have set up such a strong towncircle. Lucianos also killed the worst possible kill targets.

I think many peoples' other criticisms of the setup are fair, but not to the extent harped on in this thread. Town played well + Scum played poorly + setup is slightly townsided =/= MattZed is worst host ever.

I had a lot of fun so thanks to everyone for playing and MattZed for hosting.

[Okabe]

No you weren't. You guys couldn't win together and you only had two dead townies at that point iirc.

considering we were literally trolling the whole time[double tailor and you didn't lynch any of them], it proves the point of it being winnable. You guys were focusing on none of the scum but Sen. If scum actually cooperate in this setup its winnable. The problem is scum always try to kill each other.

[Frozen Angel]

it has happened before on offsite games. ive seen a 1v1v1 game too. know what happened? town would not vote and casues a no lynch, scums cross killed an town won.

you of all people should know that if it can happen it can happen not matter how unlikely. thats how mafia is. when you are arguing worst case scnarios you are arguing something that almost never happens so trying to apply it here is double standard

nah in MS I saw multiple mods conlusding their game then

and scums will kill the town ending up in 1v1 the other day.

its just too lame to continue playing a game like that. but thats not the reason I have issues with multi faction games. a scum should manipulate and town should solve a mystery. when you force scum to actually solve another mystery and put their lives in danger - like the way town are but they normally don't care cause there are a lot of them, your making the game extremly harder for mafia. One might disagree saying a scum who must play as a town is another type of town and playing like that is easier. I beg to differ cause they don't have the majority so they still need to be manipulative the same way another scums should. I was researching on this and the result was extremly suggesting the multiball games - unless when town has no majority overall has something around 73% town winrate (the experiment was on MS) , 10 percent a single scum faction winrate , 7 percent 3party wnrate (which is wierd cause playing solo should be harder - but its not thanks to the mess of multiball games and the rest 10 percents was the ties. Here the lower deadline actually makes the game more action analytical than Interaction anlytical so I'm sure the reults are very different - and harder to track (because I saw how your multiple faction circle thingi is)

[BananaCucho]

considering we were literally trolling the whole time[double tailor and you didn't lynch any of them], it proves the point of it being winnable. You guys were focusing on none of the scum but Sen. If scum actually cooperate in this setup its winnable. The problem is scum always try to kill each other.

Fire tried to draw attention off of Frozen, he knew her right away from FBI chat

Got him executed.

You were saying?

[Frozen Angel]

I suck as a disguised whatever. That just never works cause I have certain personality aspects and my English sucks.

[BananaCucho]

I suck as a disguised whatever. That just never works cause I have certain personality aspects and my English sucks.

You caught onto Yuki right? I haven't read the chat I just have it off of word of mouth from fire

[Frozen Angel]

actually I cought Fire - I wasn't sure about yuki

[Cancer]

=)

[Gyrlander]

=)

You guys seriously don't learn lol

[Yukitaka Oni]

=)

actually I cought Fire - I wasn't sure about yuki

You caught onto Yuki right? I haven't read the chat I just have it off of word of mouth from fire

You guys seriously don't learn lol

Day 2 chat: Frozen, you're melt ^)o.o)> (hidden meaning: you're exposed!)

[Okabe]

Fire tried to draw attention off of Frozen, he knew her right away from FBI chat

Got him executed.

You were saying?

Im talking about mafia wars one and how the setup is fairly winnable for scum?

[yzb25]

Well, any jokes aside, I think this setup is a good one. Balances problems are not that massively important IMO. Every setup can be improved after every run and this is not the exception.

This. Asymmetry is typically more interesting than fair setups, anyway. As long as each side has some potential to win, it's pretty fun.