Chungking Espresso

LARP field study: Mafia

Posted in Game Analysis, Projects by Simon Ferrari on October 26, 2009


For my LARP field study I played a night full of Mafia with Paul, Pauline, and Jenifer from class (along with a number of their friends). Doug Wilson of IT Copenhagen calls Mafia “the most political game ever conceived.” The game is an ideal LARP for non-traditional roleplayers, because there are no combat rules to remember or stats to track. Typically the game is played with between 10 and 20 people, seated in a circle. We had ten for our session, a number which lends itself to a more intimate and competitive experience.

One player takes the role of the narrator (game master) who randomly doles out roles at the beginning of each play experience, tracks the state of the game, and provides a narrative context for every game action. There are two cycles in the game: night and day. The game begins at night, with all heads bowed. Six players were assigned the role of basic townsperson; they have no special abilities or duties. Two players constitute the Mafia, and each night they raise their heads to select one person to kill. One player is the detective, and each night they can point to one person, asking the narrator if that player is in the Mafia. Finally, one player is the doctor, able to select one person per night for protection. Nobody knows what role the other players bear.

During the day stage, the results of the Mafia’s activities are reported. If the marked player was not protected by the doctor, they die. If the detective accurately discerned a Mafia member, she may want to declare the fact. But if she reveals her identity, she becomes an easy target for the Mafia if the doctor is unable or unwilling to protect her. Then the townspeople begin accusing each other of being in the Mafia, stating their (usually tenuous) reasons for believing so. Players can choose not to condemn anyone, but usually the Mafia players will attempt to sway the townspeople toward killing each other (which leads to counter-accusations, etc.). An accused player gives a defense speech, then the players vote on which person to lynch.

When the Mafia murder somebody, the narrator does not reveal what role the dead player bore; however, when the townspeople lynch a player they are told what role the dying player held. The game ends when either all townspeople or all Mafia members are killed.

It took awhile for us to get the game started. During the first round, I forgot which role I had been given and ruined everything. Everybody forgave me when the narrator forgot what was going on during the second round and spoiled that one. The third attempt was a success, especially for me. Because I knew what roles everybody had been assigned during the first two unsuccessful attempts, I used fuzzy math to try to discern which players were the most likely to be Mafia. Basically I went on the false mathematical assumption that the chance of three successive “heads” in a game of coin-flip is 1/8 instead of 1/2 (I still want a look at the theorem that establishes that bit of nonsense).

As it turned out, my fuzzy math worked! I successfully picked the two Mafia even though I was only playing a lowly citizen. The first time I nominated one of the suspect players, nobody believed me and didn’t vote for him to die. So during the next round, I falsely stated that I was the detective and that I knew the second suspect was mafioso. The healer was dead at this point, so I knew I would be killed after the round was over. I gave an impassioned speech about self-sacrifice, everybody bought it, and we lynched the suspect player. I was right about the pick, and I was also right that the remaining Mafia player would off me that night. But the real detective was still alive, and he found out who the second murderer was in time to win the round for the townspeople.

The next round, I was killed straightaway. I assume it was because I had such good hunches during the first game. This is similar to the experiment of iterated prisoner’s dilemma in game theory, where bias from previous plays affects how the players within the dilemma choose in subsequent rounds. I watched the players to figure out if any of them had tells, and I discovered that one of the players giggled whenever he was in the Mafia. During the third game, I heard the distinctive giggle on the first night and outed him to everyone during the day. After I explained my reasoning, a few players believed me and we successfully lynched him. Then I got killed the next round. Playing Mafia too well usually means you’re going to get axed.

By the fifth and final match, I’d consumed a bit too much alcohol for my own good. This resulted in me persuading the townspeople to murder two innocents in a row. I’m glad we stopped after that round. So I’ve played Mafia twice now, and I’ve never actually gotten to be in the Mafia. As a result of this, I can’t speak for how to strategize a defense while playing one of them. The rounds that I was the healer and the detective were the rounds where I died the first day, so I also don’t know how to play as those roles. Mostly I’m good at playing a standard townsperson, and I’ve got a knack for picking at least one of the Mafia off before getting slaughtered the following night (healers tend to be very stupid; they never protect me, their star player).

Is there a difference in embodiment while playing something like Mafia over a videogame? I don’t believe so. Identification with avatars in first- and third-person camera views has been well-documented. There’s a palpable, giddy energy to live action play, but for calculating players such as myself the difference seems negligible. This is probably because of the principle Gee calls the “psychosocial moratorium,” or what Huizinga calls “the magic circle”; this is a protection from real-life consequences and harm that some believe is intrinsic to play (perhaps the only exception would be in what Caillois identifies as Ilinx, or “vertigo,” play… there is a real danger present with things like roller coasters and skydiving).

I have no problem sacrificing myself for the team in Mafia, because I know I’m not dying in real life. The act of taking on a role is always a necessary step away from absolute embodiment and identification. I shun anonymity in online play, so I’m always just playing an accentuated fraction of my real self when I play any game. This appears to hold true in live play: I was sarcastic, calm, and reasonable (except when I became inebriated… which can affect performance in online games as well).

As for the strategic difference between NPCs and real human players, I hold, along with Jason Rohrer, that there isn’t much of one. I didn’t know any of my fellow Mafia players exceedingly well, so I tested and prodded them much as I would an alien computer intelligence. As a material and physical determinist, I think people behave with predictable regularity (except in panic situations). I read the one player’s giggle-tell much as I would a sound cue in a videogame. If I’d been playing with family or close friends, this might have been different–but only because I would know them and their personal rulesets all the better. They could act to upset my predictions, but I would probably be able to counter-predict that if I were playing carefully enough.

One notable exception to this rule was that we had a player named Akido who spoke little English. His defense was always, “Why do you think I’m in the Mafia? I am innocent!” It was impossible to read him, because he wasn’t fluent enough to craft different responses based on his current role and situation. I correctly identified him by luck during the first round, but every time after that (if he were mafioso) nobody was able to nail him. We avoided accusing him, perhaps out of fear that we would be discriminating against him. I wonder how this could be simulated in an NPC?

Jenifer made two videos of the experience, but I can’t speak to their quality because I don’t want to download them:
Tagged with: ,

7 Responses

Subscribe to comments with RSS.

  1. Max said, on October 26, 2009 at 11:56 am

    Interesting post – wish I could have participated! As for the coin-flipping, it’s really not ‘theoretical’ – all it is is that each flip of the coin is an independent event. It’s not influenced by previous coin-flips – how would they influence future coin flips? Each one is an independent event, and each one has a probability of 50%. It’s tempting to think “But, three in a row! How likely is THAT?” but coin flips are not grouped in any context other than our minds, and (unfortunately enough) our minds cannot influence real-world probabilities. If you toss heads twice in a row, your third throw is not affected by those previous two tosses – it’s just another toss.

    As for mafia strategy, I always found that it helped to lay low for the first couple of rounds observing the other players. Use this time to sense who will be weak to accusation. For instance, maybe another player is always quick to accuse someone of being mafia every round. After a couple of rounds, it’s VERY easy to say “Why is this guy so eager to accuse other people? I bet he’s just mafia and trying to deflect our attention from himself!” Also, if at all possible, avoid directly accusing people, and instead simply point out evidence in a leading manner that will make your fellow players likely to make the accusation that you are trying to manipulate them into. In case you can’t tell, I derive GREAT pleasure from being a member of the mafia – in fact, I don’t have much fun in the game if I’m not. For some reason, I intuitively grok effective mafia strategies much better than those that are effective for other roles.

    Also – how many variants are there on the ruleset you presented? Do you have any data for the most common variants? I know that I had never heard of a “doctor” or “protector” role before, and usually played with a slightly larger mafia than you described.

    • Max said, on October 26, 2009 at 11:58 am

      BTW, in case I wasn’t clear enough, in the second paragraph there I’m talking specifically about strategy when you are a member of the in-game mafia.

    • Simon Ferrari said, on October 26, 2009 at 12:19 pm

      I donno about variants; I remember seeing the doctor role once when I was looking through a Wiki on the game.

      I was a very proactive player, because I kept expecting the healer to protect me. So right away when day began I made my accusation and always had a good reason for it (because we allowed OOC reasoning, which I typically delivered in character… which was drunk, cynical me). Once somebody said in defense, “Why was Simon so quick to accuse me? He jumped on me right away” And I cut him off with a, “I’ve been the first to accuse every time. This isn’t a change in character.” It helped that I was never a Mafia person.

  2. TheGameCritique said, on October 26, 2009 at 12:09 pm

    There are variation on the game that only change the premiss, but I find them more interesting, because the narrator can get creative with the story. The variation I like best is Servent of Cthulu. It’s the exact same game with the roles simply renamed.

    One question though: we ususally played with only 1 killer rather than two. Would that solely be an effect of the group size or is it a personal choice? It seems to me with three people dying each round would end the game very quickly in a group of only 7.

    • Simon Ferrari said, on October 26, 2009 at 12:16 pm

      Ah, so, the two Mafia members have to decide on one person to kill. And the townspeople don’t necessarily have to lynch anyone. So between 0 and 2 people died (zero if the healer made a save).

      I’ve played a fictive variation where the Mafia were werewolves, and that was good for the acting and narration. Also played a 25-player match with three competing Mafia teams (Yakuza, Triads, and Italian). That was super ridiculous.

      • qrter said, on January 3, 2010 at 7:32 pm

        The version of this game I’ve played has been with werewolves, like you mention.

        It’s been years ago, but I really enjoyed it a lot.

        I never would call it a LARP though, more a (complicated) parlour game.

  3. Nick LaLone said, on October 26, 2009 at 8:15 pm

    This reminds me of The Chronicles of Amber diceless RPG:

    It is quite possibly the most interesting game i’ve seen work but it was probably the players I played with and the fact that our family intrigue plot worked in a way that not even the GM had thought of beforehand.

    I would guess that this probably has a prereq of reading the first Chronicles of Amber game but you could also watch any show with a rich powerful father who disappears with a myriad of semi-interested sons and daughters and get the same effect.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: