Before I break these down and delve into a case study for each (that you can play along with!), consider this thought experiment.
Someone signs up for two mafia games on two different sites a month apart. Each game has a player list with the exact same number of Village Idiots, good players, newbes, etc., and the game plays out in precisely the same way. The bandwagons develop on the same player archetypes and even the cases match post for post. The only difference is that on his home site, the player rolls town, and on mafiascum, he rolls scum. So here's the hypothesis: this player could copy and paste every single post he made as town into his scum game, and no one could tell if he's town or scum.
Theoretically, there can exist a "perfect scum game", where not a single mistake is made, whether the others players lynch correctly or not is irrelevant. There's no credit or blame for lynching this slot, the outcome is completely random, like writing a case for Groot with his post restriction. No meaningful scumhunting could change the odds, just like no behavioral cues could be gleaned from a cheating player who didn't read his role PM.
So exactly how does an experienced scumhunter stack the odds for town? This is the question this guide will explain.
To outWIFOM the scum
All I have to do is divine from what I know of you: are you the sort of man who would put the poison into his own goblet or his enemy's?
It's possible to beat the average outcomes by outguessing scum in a game of WIFOM. We're not playing a simple game of rock, paper, scissors. We're playing a long, complex game, using our mind and emotions, which is easily susceptible to cognitive bias and behavioral patterns. The probability that a player will choose a faster method of achieving a goal is different than the probability of him choosing a more cautious approach. Even without any meta, it's possible to develop a sense of what play your opponent is thinking of and react to that.
Let's say we benchmark the chance of lynching scum correctly Day 1 at 30%. Now let's say, that without looking into who the scum are specifically, you're able to guess which player the scum are most likely to bandwdagon. It could be the Village Idiot, or it could be a newbe, or an arrogant fool. Whatever the most likely. Now all you have to do is scumhunt inside this bandwagon to improve your odds of lynching scum. Using this method, you might catch scum with odds far greater than 30%. That's what I call outWIFOMing the scum. You exploit "common sense" and go as many layers deep into WIFOM as you think the scum will go in order to corner them, and lynch an approximation of where they should slide in on the bell curve of expected collective scum behavior. You target the mean and ignore the edge cases.
To Solve the Game
With a sufficiently large dataset, instead of guessing what the most probable posts and actions coming from scum are, you can tell a coherent story from start to finish about how an individual scum intended to achieve their win condition. This is a precision strike, guessing what a player thought, how that thought manifested in his strategy, which guided his actions. You weave a narrative of his evolving motives, and explain each step he took in his journey. This is where a scumhunter differentiates himself in his ability to write a coherent case.
Getting scum lynched
It's not enough to know who scum are, you have to then convince a majority of players to focus on them and get them lynched. Getting a town consensus is far from an easily repeatable task.
Be careful to not click on reveal spoilers in the 1st post if you want to play along with this guide.
SETUP: 10 vanilla town, 3 mafia goons, and mafia will only be allowed to make one nightkill on the first night; it then becomes a nightless. Each day, the mafia decides to place every player in one of three groups. Town then votes for the groups during the day, and upon achieving majority vote for one of the groups, all the players inside of it are lynched.
Can you outWIFOM the scum?
I'm certain that you can do it too now! Test your knowledge, and click on the spoilers below.
Here are my private thoughts coming into Day 1:
Spoiler:
Scum are going to create at least one group without any scum in it. That group is unlikely to be group 1. I need to lynch group 1, whether there are 2 scum in it or 3. If I publicly overshoot and tell the town I think there are 3 scum in group 1, they might be comfortable with the theory that there are at least 2 scum in group 1, and agree with me to lynch group 1.
Here are my posts attempting to outwit scum on a macro level, without any specific scumhunting:
Spoiler:
In post 73, Albert B. Rampage wrote:Scum would just put all 3 in group 1 knowing that nobody will self-lynch, and then sweep the game from there.
Here are what scum discussed in their PT in pre-game:
Spoiler:
Scum 1 wrote:I'm kind of leaning towards dividing us up somehow as 2:1:0 so we have a chance to not lose any of us, we'll just have to make sure the group with 2 doesn't get lynched
Scum 2 wrote:I think we should go 2:1:0 or gamble on 3:0:0.
I kind of like this idea: one large group with 2 scum. 1 in the smallest group. and 0 in a medium group. I think players will be inclined to go for the medium group and assume there is at least 1 scum in it.
Okay, so I'm able to convince a majority of players, and we find town consensus to lynch the group I want.
It's a 3:1 nightless now and scum are feeling confident because Day 1 lasted only 29 pages and there was almost no scumhunting done. That's where they are wrong, because I switch into solving the game.
Can you Solve the last scum?
Spoiler:
In my first post of Day 2, having prepared nothing overnight, I solve the game.
In post 729, Albert B. Rampage wrote:Okay, let's look at what we know and attempt to find answers to what we don't. Let me begin.
1) We know that mafia checked the previous game with 1/1/1 distribution from Ame's post. They chose to do a 2/1/0 or a 2/0/1 distribution so they had a
lynchbait option where no scum dies
. The question is, was that group 2 or group 3?
Of course, we know my theory:
In post 378, Albert B. Rampage wrote:
Group 3 is the most easy to lynch.
Group 2 is the safest option to continue the game with no clear winner.
Group 1 is where all the scum are hiding.
The first is less likely, as it proposes a slow game and would involve Scum!Albert involved, something that I find a little inconsistent considering that the appropriate form of distribution would be to place players who are vocally transparent in each of the groups, because at the same time as one scum is eliminated, players of relevance would also be, which also makes it evident that the composition {Clidd, Albert} is strongly antagonistic to this idea, which is why I imagine it to be unlikely. The second would also involve Scum!Albert, but it doesn't make a lot of sense because it would be better done with the addition of 1 ~ 2 players to group 2, something that would be more interesting in the sense of cost-benefit, especially because the group 3 does not seem to me to be a weak trio verbally, which would be characterized as a disadvantage for group 2 early in the game and would make this type of formation unfeasible. The third is plausible, considering that the camouflage of 2 members in the group with the largest number of players would be a safe move, while one of the members of group 3 would be instructed to push against the smaller group, being able to use both the pretext of Albert's existence, which is a slot with a shallow playstyle, as well as the numerical justification, considering that only 2 players will be lynched, therefore, the loss would be, theoretically, less than the lynch of group 1 ~ 3. The fourth would make sense only in the scenario where the two scums in group 3 were planning to deliver Scum!Albert via buss to gain town credibility early in the game, but the fact that they only put one more player in the group, instead of adding more players, reduce the damage done and imply a very early disadvantage for them, which might not be worth it in the long run, considering that there would eventually be speculation and suspicion about the centralization of votes in the group, which probably would not take long to lead to the inference that there was a bus in the middle of the wagon. The fifth would be possible in the scenario in which Scum!Dsj, Scum!Enomis and Scum!Mohab were able to embark on a risky gambit, but I believe that this would underestimate the cognitive capacity of group 1 and does not fit the profile of Scum!Dsj ( at least as far as I observed in our scum in common), where he would probably strongly suggest the change of composition due to the lack of security he would feel. The push, in this context, would also be group 2. The sixth seems more plausible to me than the fifth, considering that there are players like Scum!Hoctac and Scum!Ame who could build a narrative where groups 3 and 2 were the main lynchs , under the pretext that the numerical force of group 1 would make the materialization of a lynch unfeasible, and that it would be safer in the mathematical sense to choose groups 2 and 3. The seventh also makes sense, with the same push reasoning of compositions 3 and 5.
Conclusion:
compositions 3, 5, 6 and 7 make sense in my conception, considering that I can imagine the establishment of pushs on group 2 only because of the existence of Albert in it. That said, I am inclined to think that group 3 has expressive chances of having scum in its composition.
Here's the theory: If Group 3 was lynchbait, and town didn't go for it, clidd had to kill enomis because he was the only one from group 3 who sheeped me to vote group 1. Lynching group 1 was bad for scum, so clidd hopes to frame the remaining group 3 who were off-wagon to win the game on Day 2.
3) We know scum didn't want to get group 1 lynched, as evidenced by Ame's posts:
In post 517, Ame wrote:But I think going the route of minimizing casualties is better for our chances of winning, particularly because we can get more days and scumhunting out of it and will have flips to work off of. Whereas going G1 we are essentially gambling without prior knowledge of who is on the scum team and how they think. Going one of the other groups actually gives us evidence of how they think and we can work from there. Right now we are just random guessing and going in circles.
So why does Ame change her mind and vote group 1 at the end of the day? Most likely because the final scum said "I got this" and she put her faith in that player's math. Again, who's the numbers guy in the scum team? We know it's not Ame, and we know it's not Raya, so by elimination it has to be the third scum.
When we look at Raya's post, who is playing a more straightforward mafia game, she's trying to set up group 3 for a lynch as well:
In post 617, Albert B. Rampage wrote:The fact that no one from group 2 or 3 has hammered group 1 pretty much confirms my theory of minimal scum in groups 2-3.
This is a good point. Because of this I'll consider group 1 but first I need to actually find 2 or 3 people on it I actually think are scum and so far I'm mostly only seeing scummy players in group 3.
4) We know that if we mislynch group 3 today, it's instant-lose. So let's compare me and clidd.
In post 456, clidd wrote:''@Clidd, do you have a read on Albert outside of partition composition?''If you want, I can go into details, but basically it is because of the way he is playing. The aggressiveness he showed in defending the idea that group 1 is strategically feasible to be lynched seems to me an ignorant behavior that I see coming from an interpersonal trait characteristic of his profile, where he is probably frustrated/irritated by the gamestate, which makes sense within the scenario in which Town!Albert demonstrates emotional levels that would not make much sense in the Scum!Albert mentality, where he would be aware that his reasoning is fraudulent. In other words, it is my interpretation of the slot to distinguish that his expressions/reactions are within a natural spectrum of conduct, very different from what I feel about Ame, for example (which is forced).
He says it's ignorant to want to lynch group 1 but I'm town. What are the main differences between us?
A) I wanted to lynch group 1 from the start, which Raya and Ame didn't want.
B) I was on the final lynch of group 1, and clidd is not.
C) I wanted to avoid casting suspicion on clidd, so I could break the game in the final day and win. clidd wanted to avoid casting suspicion on me, so that group 3 gets lynched day 1, and that I would side with him on day 2.
5) We know that clidd heavily buddied me from day 1.
In post 204, Morning Tweet wrote:If i were setting up these groups, it would seem really obvious to me as scum that Group #3 is going to receive the most lynch pressure.
It's got an extra player over group #2, so in theory, town would think they get higher chances of hitting scum in there. This is subjective, but they didn't put any of the players that i know are good at seeming towny in there (like Cat, Ame, clidd, Kanna). Someone in group 3 admitted they usually get scumread.
And now, indeed, the majority of the game seems to be having suspicions towards group #3. Even all three group #3 members seem to suspect group #3, lmao
i do lean town on Mohab and Enomis as well. Dsjr i dont have an explicit read on yet but i havent found him scummy per se
At this point, im feeling there's 2-3 scum in group one and 0-1 scum outside
I feel like group #2 has a better chance of hitting scum, and if there is no scum outside group #1, it limits our losses a bit as well
I can understand your reasoning, but I believe that you are underestimating group 3. The pressure, in my opinion, would be in the lynch of group 2, considering the attempts to push during the day, and they only eased when I was able to participate more actively in the game, this clearly does not seem normal to me.
Especially considering that I expressed a scumread on Mohab yesterday and was scumhunting in group 3 without ever publicly suspecting clidd. Clidd has been mirroring this and suspecting Mohab since day 1 as well. All his actions make sense coming into today with his evaluation of me thus far.
7) Finally, I would say that clidd has been acting like he knows too much.
In post 668, clidd wrote:I agree about the game state indicating 2 scums on group 1, but I highly doubt that they are 3 because of the reads that I have on group 3. It would be an insult to my analysis ability to say that group 3 is entirely town, so my solve at the moment is 2-0-1.
He knows what he's doing. He sees where the game needs to go for him to win. There isn't any doubt or second thoughts in his play.
Now that I've laid out my case, you should feel confident in the fact that I am one of the top scumhunters on this site. I just completed another game where I nailed all of the scum and accurately predicted their strategy on day 2. Rest assured that it's normal for me to see the matrix. I'm not trying to bamboozle you. Even if I were, lynching clidd does not end the game if he's town. Lynching dsjstr or Mohab will.
I drop this bomb of a case on a shellshocked clidd, one of the townies votes for me, another one votes with me, clidd votes for me, we go back and forth for 10 pages, and in the end, we lynch the final scum, clidd, who played another excellent game (he easily won games as mafia previously).
And that's how I scumhunt by outWIFOMing the scum and solving the game. Let me know if this was helpful to you!
Guard your honor. Let your reputation fall where it will. And outlive the bastards.
In post 0, Albert B. Rampage wrote:Theoretically, there can exist a "perfect scum game", where not a single mistake is made, whether the others players lynch correctly or not is irrelevant. There's no credit or blame for lynching this slot, the outcome is completely random, like writing a case for Groot with his post restriction. No meaningful scumhunting could change the odds, just like no behavioral cues could be gleaned from a cheating player who didn't read his role PM.
I just wanted to pick up on this bit in particular: I believe that perfect "playing the townie" as scum is actually counterproductive, and not a perfect scum game after all.
When playing scum, biasing things slightly in your scumteam's direction increases your victory chances by more than the slight additional chance that you get caught reduces them. (Of course, if you take this too far, the opposite happens.) So perfect scum play is slightly more scumsided than perfect town play.
(To think about this another way: the strategy of always playing the townie gives you a win chance exactly equal to the setup's EV. A perfect scum player, against imperfect town players, should be able to do substantially better than that. In most setups, scum have a nightkill, and can thus avoid the most competent town players surviving until later in the game; thus you can expect the town players to be not only imperfect, but less perfect than average.)
One other point is that if you are trying to play in a way that prevents scumhunting ever catching you, it's not enough for your dayplay to match the dayplay you would have as a townie. There's also the converse side to think about; some other player, who would have been scum if you were town, is now town. So players may be able to catch you by townhunting that player, regardless of what you do; and if your night actions don't exactly match what that player's night actions as scum would have been, that causes another potential point of divergence.
(There are some tells that rely on this sort of difference, most commonly Burden of Proficiency: if a player is sufficiently good at scumhunting, "this player should have caught the last scum by now, they didn't, thus they must be the last scum themself" can actually be valid reasoning in some circumstances, although in my experience it's more commonly applied invalidly.)
Perfect scum game is in quotes, I feel like the word perfect is meant to be interepreted as-applied here. There is a sense in which it is perfect, and if you change it to, on one site I rolled Informed Survivor, and also my opponents are legitimately better than me to where I cannot feasibly gaslight the town into bad lynches independent of how I'm perceived, then yes I think it is "perfect" Survivor play or close enough to have served a useful thought exercise.
BoP properly functioning against a copypaster is an interesting thing and worth taking a note of, although it's a meta characteristic operating on the player as they enter the game and can be divorced from criticism of the copypaster for this reason. A Burdened player is going to lose more of their scum games when they are living too long and haven't skilled at the scum, and also more of their town game where they live too long and haven't found a scum or even have found some just not enough. There's no adjustment to copypaster's approach that can regain the lost equity: bussing even hard loses equity from the lost teammates you've bussed, bussing too little is exactly the problem.
"Let us say that you are right and there are two worlds. How much, then, is this 'other world' worth to you? What do you have there that you do not have here? Money? Power? Something worth causing the prince so much pain for?'"
"Well, I..."
"What? Nothing? You would make the prince suffer over... nothing?"
