Page 8 of 11

Posted: Fri Aug 10, 2018 10:22 pm
by Gustavo
Thank you. I was told everything on the wiki was outdated. I was told a couple of players have written good posts about how to improve play and being in rusty I’d like the read them

Posted: Tue Aug 14, 2018 2:17 pm
by callforjudgement
A lot of the wiki is outdated, but not all of it.

You can check a page's history to see when the last update was.

Posted: Thu Nov 22, 2018 7:59 pm
by naturalbreadcrumbs
In a Mountainous set-up, is it really possible for Town players, through only experience and skill, to raise the chances of a correct lynch above random selection? What's the consensus on this? Is Mafia a skill game or are Mafia players really only kidding themselves about the power of scumhunting to actually find scum?

This might sound a bit fatalistic but I was curious since the wiki has this whole thing about WIFOM and how scumtells change all the time and doubts that scumtells really work reliably.

Posted: Thu Nov 22, 2018 8:37 pm
by OkaPoka
Speaking from personal experience and with zero statistics to back it up, I think that town players are more likely to lynch worse than random lynching when playing with experienced mafia players. Then again, my sample size is small and I have no data to back up my claims, its just a feeling I have.

Posted: Thu Nov 22, 2018 8:42 pm
by naturalbreadcrumbs
What do you think is the reason for less experienced players lynching worse than random?

Posted: Thu Nov 22, 2018 8:50 pm
by OkaPoka
I think the more experienced the playerlist, the more likely that town lynches town from personal experience.

Personally, I believe that there is a higher peak to playing scum than playing town, if that makes any sense. Like there is a limit to how good you can get at being able to look at anyone and just knowing they are scum (barring meta) while playing scum, there is no real limit. Scum players, especially in no investigative pr setups, are just afforded much more freedom so given enough time, they could probably create an effective persona that is essentially unlynchable. A big part of playing scum in a normal game is both getting townread while not drawing night actions, but when you don't have to care about the latter, then all you have to do is get solidly townread which isn't too hard if you put enough effort into it.

When the playerlist overall is less experienced, I think that town is more favored than scum. It takes a lot more training to be good at deception versus seeing someone be awkward and realizing they don't fit in. We've been training our entire lives to lie and tell if someone else is lying, but keeping up a lie for months is really hard without at least some experience.

But your mileage may vary and this is just all speaking from personal experience.

Posted: Thu Nov 22, 2018 11:03 pm
by callforjudgement
In post 177, naturalbreadcrumbs wrote:In a Mountainous set-up, is it really possible for Town players, through only experience and skill, to raise the chances of a correct lynch above random selection? What's the consensus on this? Is Mafia a skill game or are Mafia players really only kidding themselves about the power of scumhunting to actually find scum?

This might sound a bit fatalistic but I was curious since the wiki has this whole thing about WIFOM and how scumtells change all the time and doubts that scumtells really work reliably.
In Nightless games, town consistently do a lot better than they would via random lynching. This pattern doesn't hold in games with a regular nightkill.

The most likely conclusion from this is that although townies can raise the chance of a correct lynch above random, scum are quite good at identifying which townies are thinking along the right lines and nightkilling them.

Posted: Fri Nov 23, 2018 5:52 pm
by naturalbreadcrumbs
I see. I guess that's why town PRs are pretty vital to the game.

Maybe I should try some Nightless games after I get enough newbie games under my belt.

Posted: Tue Nov 27, 2018 4:37 am
by Persivul
In post 177, naturalbreadcrumbs wrote:In a Mountainous set-up, is it really possible for Town players, through only experience and skill, to raise the chances of a correct lynch above random selection?
D1 - no.

After that, if VCA and NKA are properly applied - yes.

Posted: Tue Nov 27, 2018 7:30 pm
by naturalbreadcrumbs
What are VCA and NKA?

Posted: Tue Nov 27, 2018 10:46 pm
by BNL
Vote count analysis and night kill analysis respectively. This means finding patterns in players’ voting history and who gets killed to find scum respectively.

Posted: Wed Nov 28, 2018 1:37 pm
by naturalbreadcrumbs
Oh, I see. So that would mean something like "player A accused/voted for player B, then A was nightkilled, so B must be scum"?

Posted: Thu Nov 29, 2018 4:27 am
by callforjudgement
That sort of reasoning is considered to be nightkill analysis, yes. Votecount analysis would be something like "player A has flipped scum, players B and C were normally very free with their votes but never voted for player A, that's suspicious".

Some people use this sort of analysis more than others. (For example, I normally check to see if someone's votes are consistent with their stated reasoning, but in shorter games, consistently voting town or consistently voting scum are both things that can easily happen to a townie by chance; so in short games, VCA is of limited use, and thus I don't really use it.)

Posted: Thu Nov 29, 2018 3:26 pm
by northsidegal
In post 183, Persivul wrote:
In post 177, naturalbreadcrumbs wrote:In a Mountainous set-up, is it really possible for Town players, through only experience and skill, to raise the chances of a correct lynch above random selection?
D1 - no.

After that, if VCA and NKA are properly applied - yes.
i fundamentally disagree with everyone who espouses the "day ones are useless" philosophy. reading people based on their play really doesn't change at all day to day, and that's not something that you can ever just ignore.

Posted: Thu Nov 29, 2018 6:03 pm
by naturalbreadcrumbs
In post 187, callforjudgement wrote:That sort of reasoning is considered to be nightkill analysis, yes. Votecount analysis would be something like "player A has flipped scum, players B and C were normally very free with their votes but never voted for player A, that's suspicious".

Some people use this sort of analysis more than others. (For example, I normally check to see if someone's votes are consistent with their stated reasoning, but in shorter games, consistently voting town or consistently voting scum are both things that can easily happen to a townie by chance; so in short games, VCA is of limited use, and thus I don't really use it.)
But doesn't that run into WIFOM problems? If players know that people who mysteriously never voted for scum are likely to be under suspicion, doesn't that mean that they'll just adjust their play so that they vote for their scum teammate earlier in the wagon?
northsidegal wrote:
In post 183, Persivul wrote:
In post 177, naturalbreadcrumbs wrote:In a Mountainous set-up, is it really possible for Town players, through only experience and skill, to raise the chances of a correct lynch above random selection?
D1 - no.

After that, if VCA and NKA are properly applied - yes.
i fundamentally disagree with everyone who espouses the "day ones are useless" philosophy. reading people based on their play really doesn't change at all day to day, and that's not something that you can ever just ignore.
I can see why analyzing day 1 dialogue in the later days would prove to be useful, but on the first day itself, unless you get lucky and hit scum with a wagon, how would you hunt for scum? There's no inherent information distinguishing them from town, right?

Posted: Thu Nov 29, 2018 6:23 pm
by northsidegal
In post 189, naturalbreadcrumbs wrote:I can see why analyzing day 1 dialogue in the later days would prove to be useful, but on the first day itself, unless you get lucky and hit scum with a wagon, how would you hunt for scum? There's no inherent information distinguishing them from town, right?
Do you believe that people lying and people telling the truth do both of those actions in the exact same way? If so, you would be factually incorrect.

In real life, at least, there are various traits of liars that can distinguish them from people telling the truth: that's the principle behind the polygraph (despite its dubious reliability) and police interrogations. This is perhaps a lot more noticeable in face to face mafia where the stress of continued lying gets to people in a way that might be a lot more obvious than could come across through a textual medium like a forum. That being said, in principle it's still visible through a forum – just because there aren't necessarily
objective
rules for what town do versus what scum do, that doesn't mean you can't analyze things within the context of the game: for example, if someone's thought process is consistent, if it seems like they have conviction in what they're saying, etc.

On something of a separate note (and I bring this up separately because I know that some people disagree with me), a tool that is almost always applicable regardless of the day phase would be using a player's meta. If someone posts a lot as town but doesn't post as much as scum and they're posting a lot in Day 1 of a game that you're in, well... that's something that points towards that person being town that you didn't need any flips to derive. That's obviously a very simple example, but there are much deeper things you could look at, all of which basically come back to the principle of "what's the difference between someone's townplay and their scumplay".


Trying to analyze everyone's alignment simply in terms of the information gained by flips is, in my opinion, almost certainly bound to get lose you games to scum who do things that you just don't expect: bussing, white flag gambits, nightkilling strange targets, etc.

Posted: Sat Dec 01, 2018 4:05 am
by callforjudgement
In post 189, naturalbreadcrumbs wrote:But doesn't that run into WIFOM problems? If players know that people who mysteriously never voted for scum are likely to be under suspicion, doesn't that mean that they'll just adjust their play so that they vote for their scum teammate earlier in the wagon?
Yes, and VCA fans try to allow for that. For example, many players who are heavily into VCA do things like guessing where scum would likely be on the wagon.

Personally I think regular scumhunting is more useful than VCA for this sort of reason, but analysing votes isn't entirely useless – sometimes when scum try to vote artificially to avoid being caught by VCA, they look scummy simply because the artificiality of their actions comes off in their posts.

On the subject of nightkilli analysis, you normally
want
to provoke scum into making weird nightkills! That means that the most threatening townies are still alive, giving town a large advantage. So even though it's WIFOM, you need to at least force the scum to consider not making the obvious kill. (As such, I'm more of a fan of nightkill analysis than voting analysis.)

Posted: Tue Jan 15, 2019 12:26 am
by BuJaber
I have never played a mountainous before, I really want to, it sounds like a dream.

I cannot believe that town would do worse than random lynching.
When all nightkills are equal (PR-wise) there is no such thing as a townie trying to avoid NK or a townie trying to bait NK. There is one single purpose shared by all townies: to find and lynch scum. We are not robots playing the game. There are certain things that just sound off that you can pick up on. Scum are also always playing with a different objective. They have to be. I can come up with a number of different strategies, and for each of them, there will eventually be a difference between this strategy and what the player would/should be doing if they were town. It just seems unavoidable.

The only way scum would win is by doing so well in looking townie they become unlynchable. In which case random lynching actually improves EV for town. However before you get to that stage you'd probably end up killing their partners with pretty high accuracy. I haven't seen anyone become a deepwolf without somehow managing to distance themselves from their partners, usually involving bussing. Basically if town had been random lynching the last scum would never have deepwolfed in the first place.

So I think that when town win mountainous they are doing better than random lynching.
And when scum win mountainous town START off doing better than random lynching then their accuracy takes a nosedive.

Posted: Tue Jan 15, 2019 10:20 am
by mastina
In post 192, BuJaber wrote:I cannot believe that town would do worse than random lynching.
When all nightkills are equal (PR-wise)
You need look no further than this to see where your premise goes wrong.

Just because a game's mountainous does not mean all nightkills are equal.

There's a huge difference between selecting Ellibereth as an N1 nightkill and, saaaaay, oh, I dunno, Not_Mafia as an N1 nightkill.

Which do you suppose would more heavily negatively impact the town?

It is precisely for that reason that towns do worse than random lynching: their most competent players, one after another, are killed off, leaving only the most incompetent town players alive. These players are the ones that are scummier, these players are the ones that are less logical, these players are the ones more prone to being manipulated, these players are the ones more prone to being in the extremes of self-doubt or confirmation bias rather than in the ideal happy medium of the two. These players are mislynch bait who frequently don't take the game too seriously--and the scum leave them around, making it easier to blend in with them, or alternatively, to become a "town" leader and take control of them.

If every flock needs a shepherd yet you kill off literally every single town player capable of leading the flock, it isn't too hard for scum to take that role of shepherd and suddenly, they sweep.

Posted: Tue Jan 15, 2019 10:58 am
by callforjudgement
In support of mastina's point: the biggest thing that determines the relationship between EV and town win rate in practice appears to be the nature of the scum's nightkill.

The more control scum have over who dies, the worse town does relative to EV. (In the extreme – a pure Nightless setup – town win rates tend to be much better than EV.)

Posted: Tue Jan 15, 2019 11:26 am
by BuJaber
Yes absolutely, I agree with that when players aren't equal in skill.

Eventually though wouldn't this become less of an issue?
To use your example, how would Ellibereth win as scum?
If such BoP exists, it makes sense that you'll end up having these same players lose as scum simply for not being killed.

Since this exists in all setups why shouldn't we assume it to be negligible and evaluate setups in their pure form as if all players were the same?

Posted: Tue Jan 15, 2019 11:36 am
by mastina
In post 195, BuJaber wrote:Eventually though wouldn't this become less of an issue? To use your example, how would Ellibereth win as scum? If such BoP exists, it makes sense that you'll end up having these same players lose as scum simply for not being killed.
Ellibereth may be the most extreme example of paragon-level town players, but even he is not infallible as town; he can as town delay his gamebreaking, he can as town end up only naming one scum, he can as town fail to name every scum as being scum, so those traits can be played up if he himself is scum--people still trust him enough to give him this leeway and only turn against him when given strong overwhelming evidence of him getting wrong what he
should
get right.

Getting a read on, say, DeasVail wrong is something people wouldn't condemn him for, whereas getting a read wrong on chesskid would be a nail in his coffin. He's perfectly capable of, as scum, manipulating circumstances where he gets reads "wrong" on players he can get wrong reads on, and gets reads right on players he can afford to be right on that he knows he would get right.

And even then
, as I alluded to--not every player is Ellibereth. He is, as said, the most extreme. If HE, the most extreme, is capable of this...then someone who is a less-stellar town player than Ellibereth but still much better as town than scum would be able to pull it off even more flawlessly.

Posted: Tue Jan 15, 2019 11:48 am
by northsidegal
In post 195, BuJaber wrote:Since this exists in all setups why shouldn't we assume it to be negligible and evaluate setups in their pure form as if all players were the same?
setups are, in one sense, evaluated as if all players were the same – that's EV. the difference between EV and actual results is the difference between the theoretical world where all players are the same and lynches are random, and reality.

Posted: Tue Jan 15, 2019 12:15 pm
by BuJaber
In post 197, northsidegal wrote:
In post 195, BuJaber wrote:Since this exists in all setups why shouldn't we assume it to be negligible and evaluate setups in their pure form as if all players were the same?
setups are, in one sense, evaluated as if all players were the same – that's EV. the difference between EV and actual results is the difference between the theoretical world where all players are the same and lynches are random, and reality.
What I am struggling to grasp is why we only have these two choices for our base assumption.

Why can't we consider a middle-of-the-way scenario where players are not equal but they are not lynching randomly?

Are we saying that the deviations from random EV are caused by the variation in skill level itself?

Posted: Tue Jan 15, 2019 12:22 pm
by northsidegal
In post 198, BuJaber wrote:Why can't we consider a middle-of-the-way scenario where players are not equal but they are not lynching randomly?
you can consider that, roughly. mathdino wrote a program that does it.

i would say that derivations from EV are caused by both. a great town can win a scumsided setup and vice versa. in the long run, if we assume that over time the number of strong towns rolled equals out with the number of strong scumteams rolled, i
think
that the results should show roughly how balanced a setup is.

i'm not an expert on game balancing really, but i'm pretty sure that's an alright definition of balance.