Back to Welcome to EM!

The Competitive Guide to Epicmafia

almost 10 years

Hello. If you are reading this, you are interested in better understanding the fundamental nature of the social game you are playing. In this thread, you will find a wealth of information on how to utilize probabilities and social dynamics to boost your game-play in Mafia and other social-based party games.



Breaking Things Down

First, a quick question we should be asking ourselves: why take Mafia seriously? This may seem like a no-brainer to a dedicated player on the site, but it's a good starting point for our discussion.

So, why play the game with any real cognitive effort? Because it is a simulation of decision-making in groups you will encounter in real life. You will, without fail, experience a real environment in which the lessons you learn in the game of Mafia will be played out with shocking similarity.

Why is this the case? Because, at its heart, Mafia is a game of communication between players who have different goals in mind. I cannot stress how important that exact phrasing is. Because EM is all about a game of communication, it is FUNDAMENTAL that you learn what aspects of communication play a part in the game. Not all are literal conversations! Let's get into that now.



Expected Payoffs

Remember, the game is conducted between players with different goals in mind. Usually, these goals are centered around the alignment or role a player receives; however, this is not always the case, and indeed, the exceptions to the rule almost overshadow the rule itself.

Because players often play with the intent to play in favor of their friends, play to lose (trolling), or simply play with little to no involvement (which often cannot be regulated, given the nature of the game), you must learn to understand just how VARIED the range of GOALS are in the social game you are playing.

> Different goals implies different payoffs and expectations for > decisions. When a player makes a choice or states something in the > game, he does so with an expectation in mind that leads to a payoff, > relative to the outcome of the action.

Let's consider an example: N1 actions. This is an indirect form of communication between players, and you, as another player, need to seek out a way to define the payoffs of the other players.

So much can be learned from the first night of interactions, before anyone has even spoken in the game. Before we go further, assume that randomization does not exist for the time being. Every action has a meaning, or so you should assume for now.

Now, the game has begun, and we have entered N1. It is Classic Mafia, there are 7 players; 2 'nilla mafia, 1 cop, 1 doctor, 4 blues.

First, let us examine the payoffs of the 2 mafia players. They wager a 2/5 chance to hit a power role, the cop or the doctor. So how should the mafia respond? They cannot, of course, know who is what role. Therefore, the 2/5 chance is not worth taking. There is a far more dominant strategy for them to undertake, and this is where we see the first conflict of expectations occur.

The doctor's 1/6 save chance is not going to aim for the cop, which would be a 1/42 (1/6 of save x 1/7 rolling cop) chance of being both the kill target and the cop. No, instead the doctor will investigate for more information on the players, if using a dominant strategy. This is why we see the more veteran and better-stats players being saved consistently in the standard setup.

Let us go back to the mafia's strategy. Their dominant strategy would normally be, of course, to hit the objectively best player in the room. But wait! The doctor's dominant choice, the one with the most payoff, is to save that person. So what now? Mafia aims for the second best player, or even the third best.

If the doctor is good enough, their expectations will anticipate the expected payoffs of the mafia in this N1 action, and they will instead choose to save the second or third 'best' player in the lobby.

As you can see, there is a conflict of expectations that leads to varied strategies in the game, which each lead to different payoffs. What is your goal as a player? To minimize the variation and insure you achieve the payoff you desire. To do this, USE EVERY PIECE OF INFORMATION THE GAME OFFERS.

So, to go back to mafia's decision here, let's say there are 2 brackets of players in the lobby. 3 players joined together and spoke with one another in the Pregame Lobby. The other 4, of which yourself and your mafia partner are included, are random players. Mafia should not target the group of 3, because doc is likely narrowing their strategy to a 1/2 save, if the doc is in that group of 3. Aim for the other 2 players in the group of 4 randoms.

But wait! What if both those players are very good - doc might be saving one of them! If this is the case, filter the group of 3 further. Check EVERYTHING the players offer you; their profiles, their stats, and anything said in-game or in pregame. EVERYTHING IS INFORMATION THAT CAN ALLOW YOU TO BETTER YOUR CHANCE OF ACHIEVING YOUR DESIRED PAYOFF.

Let's leave it at that for now. What does cop do in the meantime? Cop has a guaranteed action that no one can stop (assuming he's not n1'd). He can reliably check the 'best' player's alignment as a measure to get the player on his side (if inno) or quickly determine a threat (if guilty).



Mechanical Tactics

  • Hammering the Night - If you are a role with a night action, you should be aware of how useful a night-hammer is. If you are cop, it gives you an edge on the daystart, especially if you have copy-pasted a potential report/report format (ex. '<name> guilty/inno' '1/2'). If you are in a setup with unclear death info (ex. Fancy Pants/multisetups; janitor games where jan forgets to mention; no-reveal setups), mafia can hammer the night early to check if a PR that can hammer the night is still around. If the night doesn't end, you know a PR is alive. There are other uses; be creative.

  • Vote Blitzing - An often overlooked aspect of the game that can secure wins. While normally blitzing applies to kicking a majority vote on a player who is town, it can be used in other ways. If some town players are splitting their vote, mafia can kick and cause confusion to force a NL or even draw out a wifom blitz (eg. falsify a blitz on a mafia player in a ML situation, forcing the vote to instead go on another player). As mafia, vote-kick is your friend, and you should not hesitate to act on it.

  • (Implied) Draws - MYLO favors Mafia. If you do not need to kill on MYLO (i.e. no dangerous role or player is alive), just don't do it. If you have just killed and town NLs, it is optimal to NK and force town to lynch. Obviously, for town, if mafia NKs and you have just lynched, NL until mafia kills. Each side is forced to take action if the other has just lynched/killed. Make use of this.



Social Tactics

  • Reaction Testing - Always, always, always reaction test. If you are cop and you are uncc'd (or any position in which your reaction test will solidify your claim), fake a report on a player for reactions. Even if you have an alive inno, it is extremely useful to do this. Especially in casual games, you will see that a player who is wrongfully guiltiedd by an uncc'd cop will easily display genuine a reaction. Judging authenticity is, of course, a talent that takes practice, but remember that mafia is a game of communication between players who have different goals in mind. Guiltying a hooker that has been lawyered can give you a reaction that demonstrates they are already aware that you are reaction testing them, and that is a reaction in and of itself.

  • Roleswapping - This is another very overlooked aspect of the game. It tends to be risky if done haphazardly, so let's examine a very specific aspect of roleswapping. Consider a situation in which doctor has died in our Classic Mafia game, and cop is hiding. Of course, meta has the game get NL'd out of D1. However, if you are a blue, you have a strong chance to 'hypo' cop. Don't out a report. Instead, 'soft' a report by hammering the day with a vote on that player. If you want to be more obvious, ready yourself with a guilty statement and enter it into chat just as you hammer the vote on that player. This is a very specific instance of roleswapping which is highly effective in ranked games. It doesn't matter if you guilty an innocent player, because you can retract the claim and generally no harm will come of it. This instance of roleswapping also utilizes a town-sided mechanical functionality of 'blitzing' or hammering. In general, standard roleswapping should be done through whispers, but in the absence of whispers (current comp meta), you can reliably do it with friends you normally play with. Ascertain alignment, and then act accordingly.

  • (Reverse) WIFOM - WIFOM is a fairly misunderstood concept. It's better understood through the application of expected payoffs. In anticipating another player's behavior, acting in an opposite or variable fashion can often provoke an uncharacteristic response. In doing so, your ultimate goal is to return to your dominant strategy while misleading the player you are WIFOM'ing into behaving in a non-dominant fashion. For example, let's assume doc can either save a previous save or the uncc'd cop in our Classic Mafia game. 7 people are alive, D1. Doc's optimal play would be to mislead the mafia into thinking that they will not save the cop, but doc should ALWAYS save the cop in this scenario, because this is the choice with the highest expected payoff. Remember to anticipate expectations and how they alter payoffs, and WIFOM will make more sense.

  • Tunneling A very characteristic scenario in Mafia is the FOS (finger of suspicion) that seems to never go away. If you are being tunneled by a player, the best way to appeal against it is to understand their goal and demonstrate that you share their goal. This may even be trolling the game. What matters is that you convince them you are on their side. One thing you can learn from an obstinate player that is well-analyzed and persuaded is that there is a high probability that they are not actually town-sided if they refuse to budge. OMGUS is an overused term, but it is a real thing. On the flipside, tunneling is a very useful tactic if you are unsure of your reads in the game. Pressing a player to respond, in any fashion, will provoke behavior that can lead to more certain evaluations of their goals (and from there, their alignment).



Reactions to Key Events

This is perhaps the most neglected factor in all of EM. Let's list a few critical moments in the game: Day-start, Shot Fired, The Last Minute before kicks, Day-end, Claims, Doc-saves/other no-kill scenarios, and the Interim.

Day-start can show you who is present immediately after the night. D1 usually also offers the first line of speech for the player. The first line is always critical, because it frames the context through which the player identifies with the game. It is a critical window into their expected payoffs that should be reconciled in your perception of them.

Many players tend to dismiss last-second messages given at the end of day. I urge you to anticipate very telling statements at day-end. Often, if there is a CC, a player's last few statements can be peered into unlike other ATE (appeals to emotion) they have offered. You want to be ready to catch these final statements, as they can be the most revealing and unadulterated examples of their persona and perceived payoffs.

Counterclaims are a little less obvious than a shot being fired or a doc-save/equivalent scenario. Timing of claims has little to do with authenticity. In fact, I encourage you to neglect the claims temporarily and instead analyze the players analyzing the claims. After you have seen how others react to the claims, then you assess them for yourself. Note any alliances being formed around a player, and try to understand why that may be occurring. Often the most revealing tell of a claim is who is defending them, regardless of how well the person claiming performed. This is because the ones defending the claim are less aware of their own presence while dealing with the claims, and mafia often consider this an opportunity to embolden themselves, assuming the given that they are not being lynched that day.

A vastly invisible yet consistently prevalent phenomenon is the middle of the day interim period. I call this the Interim. It doesn't necessarily happen in the middle of the day, but it almost always happens at least once in a game. At some point during the game, a sudden absence of discussion will follow; you might call it an awkward silence. If you are vigilant throughout the duration of the game, you will catch this moment and notice who is talking during it. This will provide you with invaluable information about who is really interested in the outcome of the game, as well as why they are interested (if you've been reading them well thus far). Look out for the Interim. It can tell you much about who is really lurking and who is in constant-discussion mode.



The Scumhunt

There is a very simplistic way to view scumhunting, and we have already went over it, as Mafia is a game of communication between players who have different goals in mind. Observe a player not for objective indications of towniness or scuminess; this will delude you into assumptions that players who are town play for the sake of winning. Assume that every player has a different modus operandi, with each seeking a different outcome from a different set of payoffs.

Think of Mafia as a sort of story, and the characters in the narrative have all been given free reign to create their own part of the story. Each character slowly offers a portrait of itself as the story unravels and the game progresses. Observe and catch every instance in which a character betrays the self-portrait they have created. If you have conceptualized the person well enough, you will know when they have betrayed themselves to a purpose that does not necessarily fit with the framework they have established for themselves. We'll call these group of players the divergent type.

Alternatively, another way to observe the game is essential if you want to catch the convergent types that are mafia. These are well-trained players that create personas that are, of themselves, demonstrating payoffs and expectations that are in line with town winning the game. However, they turn out to be mafia! How do we catch those types?

Well, one method is to break down their character and force them to diverge from their persona. This can be achieved through traps, which vary from WIFOM to roleswapping to simply tunneling. Ply the player into acting unnaturally and see if that demonstrates a divergence or a convergence on their frame of narrative.

Another method to define a convergent persona is to become one yourself. Of course, this is as good a method as any to scumhunting, as you put yourself in the other player's shoes. However, the idea here is specifically to create a bond between you and the player, so that they assume trust in you and you trust in them. Basically, you force a new framing of the narrative, such that they must interact with you or compromise themselves. This may seem very esoteric, but if organized correctly, a bond of trust (or possibly, hatred - see tunneling) can alter the entire progression of the game, such that you can alter the framework through which other players are expecting their payoffs from actions/communications.



The Towntelling Mafia

There is a lot to be said about how differently you can play the mafia alignment, so I'll briefly describe one of the more consistently successful ones. In this setting, you assume yourself to be a town player. Act entirely as if you are trying to win for town. For novices, I suggest covering your roles and even having someone else play during the night phases. For experts, you should be able to disassociate from those phases and completely configure yourself during the day to be 100% in accordance with town behavior. This means offering strategic help, lynching your teammates if need be, leading the town with useful suggestions, and critical analysis of every player. Just scumhunt, champ, and your persona will be converging towards town victory. Just don't get caught up and reveal your expectations/payoffs; be unpredictable if necessary.



Meta-Gaming

Meta is a huge component to any social game. It's impossible to prevent friends playing together, and it's even more improbable to avoid players utilizing past games as indications of current behavior.

One key thing to remember when dealing with Mafia's meta, however, is that your own meta should never be too revealing, and other players' meta is always circumstantial unless you yourself have witnessed that behavior. Even then, proceed with caution: meta is one of the least reliable indicators of a player's goals/expected payoffs. That, of course, can be to your advantage, if you frame the meta (or, your site-wide persona) to be such that you are perceived as someone always worth keeping alive.

Remember: every piece of information available on the site can help (de)construct the character you play as. Stats (and thus win-rate/trophies), for example, play a HUGE part in solidifying a player as trustworthy or worth keeping alive.



Some Final Thoughts

If you want to win at a social game, it's important to remember that leadership is a key communication skill you will have to rely on. Leadership is a varying talent, and different aspects of your personality can shine through in different ways, charming different people (...game of communication between players who have different goals in mind).

But winning isn't everything. Enjoying Mafia has little if anything to do with victory. If you want the most out of your Mafia experience, you will try to play around with what has been suggested here, to create an enriching playground for your intellect and your own behavior. Test the fields and see what works, what doesn't, and what will give you certain results.

Mafia is a social-based party game. Don't abuse the fun and turn it into a torturing trial-by-conversation for yourself. Flourish through your mistakes and your unique attempts, because you can learn a lot about life from this game.

If you've made it this far, thanks for reading, and I wish you the best of luck in all your pursuits.

I may add on more to this thread in the future. Feel free to discuss/debate/argue with what I've posted, and I am more than happy to accommodate different conceptions of the somewhat jargon-heavy ideology put forth in this thread, as well as contentions.

Cheers, Vex

deletedabout 8 years

Arcbell says

i find it funny that in doing this pranay ur actually trying to make relevant posts in the topics


can you tell me how to win more...i don't win enough...and i'm a good player...we all know it....i think.....

l o l

while you're at it help that panthero guy out, he's bad
about 8 years
I completely had no idea that something like this existed.
about 8 years
i find it funny that in doing this pranay ur actually trying to make relevant posts in the topics
about 8 years
some people need this
over 8 years
I saved this link on a Notepad doc, extremely helpful, cheers.
over 8 years
This is a good guide. A critical point I'd bring up would be that overuse of reaction tests can lead to a too-predictable meta for yourself while at the same time not accomplishing much in high level play because mafia playing there can easily fake genuine reactions to all but the most complex reaction tests.
deletedover 8 years
This is a good thread, it deserves to be stickied IMO.
We alreadyseem to have a competitive guide stickied tho
over 8 years
bumping this cos I am bored and wanted to necro a topic, and foundd a legitimate topic for it:D
almost 10 years
Incredible thread, enjoyed reading every word of it.
deletedalmost 10 years
People in comp need to learn to grow a goddamn bloody freaking spine.
almost 10 years
Another thing worth mentioning about expected payoffs:

Let's say you are Cop, it's N1, Fancy Pants. If someone in the game is an individual you expect to kill you, since there is no reliable chance doc can save you, if that player is mafia, you have a strong chance of dying. Thus, checking them for a report is probably not your best option.

Because they are likely targeting you (perhaps even expecting you to check them if you were cop), your expectations about your payoff changes, as it is now unfavorable to go after who might normally be your optimal choice.

So basically, you can safely eliminate checking a player who is -always- going to kill you, because the report you would get on them would not matter if you're already dead. This logic can be applied also to doc saves, etc. so forth. Consider the role action and what your friend/enemy is likely going to do to you/someone else.

Thus, through analysis of expected payoffs, you can narrow down who you should be checking N1, even.
almost 10 years
Another thing worth commenting on is tonality in the game. If you are able to pick up on tones, you can quickly notice any divergent tendencies in a player's persona in the game. Tones vary greatly and often unavoidably in competitive play. This may seem obvious at times when reading an emotional, towntelling player, but it can be more subtle when a mafia suddenly decides to fence-sit (sits between two different FoSes, typically at 3-way lylo, in order to gain favor of one player over the other) and you pressure them on that fence-sitting.

Happy Holidays everyone!
almost 10 years

Jaleb says

I disagree with your idea on reaction tests. It seems poor if you get a bad player to show weakness and out some pr (doc/watch/bg) to see if you think they're town. Big cases would be in B2NS, and BV where the reports are poor except for on d2 anyway. Knowing this fact, some players might have the same reaction to any report on them. If this was face-to-face mafia, try it. Since this is forum, it's usually better to get reads other ways.


You're right. There are definitely important caveats to reaction testing, and I meant to edit out the "always" part, but sadly, it doesn't let me edit the OP.

But yes, reaction tests are very useful if those situations aren't arising (eg. prs are already out or are going to out that day). So yeah, be wary of using it, but I'd still say give it a shot when it doesn't endanger town's other PRs. As for false innocents/guilties, which I'm not sure you if you were also referring to, I have to say I still think it's a great idea to test on, say, a Godfather/Lawyer setup.

One thing I neglected to mention that should really be included up in the OP is the actual kill behavior/sequence throughout the game. It's a fairly underutilized pattern that a player can observe and ascertain why the mafia killed in that order/who they killed/etc. Kill choices and timings, as well as the behavior of the remaining players' response to those kills, can be extremely powerful reading tools for a keen player.
almost 10 years
I disagree with your idea on reaction tests. It seems poor if you get a bad player to show weakness and out some pr (doc/watch/bg) to see if you think they're town. Big cases would be in B2NS, and BV where the reports are poor except for on d2 anyway. Knowing this fact, some players might have the same reaction to any report on them. If this was face-to-face mafia, try it. Since this is forum, it's usually better to get reads other ways.
almost 10 years
Tl;dr
almost 10 years
Good.
almost 10 years
I havent read it yet but it looks long so I plussed for effort even if it's secretly 2nd world war germany (lol @ censor) propaganda.
deletedalmost 10 years
Overall logical, insightful for noobies, lets people know the necessary game mechanics. final judgment: 0/10 would not bang