Normally, I would check this out. But I’m going to be traveling for a couple of weeks in September and expect to have limited online access at best. So I don’t want to get involved in a game I won’t be able to keep up with.
You could just check out the board, read some stuff if you like. We’ll see ya when you are ready
Another newb here: What are some of the basic strategies? Does anyone know what percentage of the time the scum would win (assuming they represent 25% of the population) if people were lynched completely at random? Thanks.
The two key factors of the game are that the villains will be killing 100% villagers and that the villagers have greater numbers.
The problem with that situation you described is that the villagers will be mostly killing other villagers (75% of the people chosen at random are villagers) and the villains are targeting villagers (so 100% of the people they choose will be villagers). So in a game where 75% of the players are villagers, 87.5% of the players who get killed will be villagers. In other words there are three times as many villagers but they’re being killed seven times as fast.
If the percentage of villains is 25%, the break even point is when the villagers are killing villains with a 50% accuracy rate. That gives you a villager kill rate that is exactly three times the villain kill rate.
ETA: Upon further reflection, the above figures are not accurate. The percentage of villagers starts at 75% but will lower during the game due to the targeting of the villagers by the villains. That means the chances of a villain being killed at random will rise as the overall percentage of villains rise. But I’m not going to do the calculations.
It doesn’t depend on the percentage. From the cases I’ve calculated, I think a balanced game requires the number of Scum to scale as the log of the number of Town. Thus, for instance, a vanilla game with 2 Scum and 8 Town is almost exactly 50-50, but double the numbers to get 4 Scum and 16 Town, and it’s an overwhelming advantage for Scum (something like 80 or 90%). For a vanilla game to be balanced with four Scum, you end up needing something like a hundred Townies (which is of course completely unworkable).
OK, I have time to run the numbers now, and they’re not quite as extreme as I remembered. Assuming that Scum need to outnumber Town to win, for 8 Town to 2 Scum, Town has a 50.78% chance of winning. For 16 Town to 4 Scum, Town has only a 33.23% chance. With 3 Scum, you need 19 Townies for a balanced game (50.34% Town win), with 4 Scum, you need 34 Town (49.96%), and with 5 Scum, you need 53 or 55 Town (49.66% or 50.44%). Note, by the way, that Town’s chance of winning drops significantly if the total number of players is odd: With 54 Town vs. 5 Scum, for instance, Town only has a 35.56% chance of winning.
Thanks, that’s interesting. I assume that, as the Wiki article suggests, townies do somewhat better than those numbers in practice (since they can look at the voting records to help identify more likely suspects), yes? Although, OTOH, if the game has only 8 or 10 players, I’d guess there wouldn’t be enough rounds of voting for that to amount to a large advantage.
That depends. The Scum can, if they so choose, behave exactly as if they were Townies, making no particular effort to try for mislynches and sanguinely (an apt adverb if ever there was one) accepting the loss of their teammates when it happens, in which case there’d be nothing to find in the voting record, and the game would play out exactly as though lynches were completely random. Alternately, they can choose to put their thumb on the scales and coordinate their voting somehow, which probably helps them on any individual lynch in the short run, but provides useful data to the Town in the long run. The fact that the Scum can make this choice, but the Town can’t make it for them, gives the Scum an edge, but quantifying that edge is extremely difficult, and would require a lot of very messy game theory.
Thus Town should have a slight advantage in a perfectly random game, correct?
I’ve often wondered about this stuff, but have been too chicken to ask, as I’m unlikely to be ready to play anytime soon. I find it hard not to be obsessed, and I can’t take 6 weeks to a month being obsessed over a game.
I always miss these threads when they start. I don’t have anything new to add other than that the best way to learn is to play. You don’t really “get” the human element of the game until you play.
No, because there’s two rounds of voting per day. The Town votes as a whole (villagers and villains) on who to kill each day. But the villains alone get together and vote to kill somebody each night. And the night vote isn’t random - all of the villains know who they are. So half of the people who are killed (the people killed at night) will be villagers. The other half (the people killed during the day) will be a mix of villagers and villains. This imbalance gives the villains a significant advantage. The villagers have to try to overcome this imbalance by figuring out who the villains are and raising the amount of villains who are killed in the daytime.
To do this, the villagers need information. They have to track how people voted and figure out who’s voting in a pattern that favors the villains. The villains have to figure out how to influence the day’s voting without the villagers figuring out what they’re doing.
What Chronos described is an alternative strategy where the villains don’t try to influence the day votes. They all act like genuine villagers and don’t use the knowledge they have as villains. If they succeed in this they become invisible and the villagers have no way of identifying the villains. The people who are killed during the day become a random mix of villagers and villains. The villains meanwhile retain their ability to target villagers at night.
This looks correct, assuming that by “log” you mean the function called “square root” in English-speaking countries. 
Let t be the total number of players and s the number of scum achieving some specific win rate, e.g. 0.5. After one cycle, 2 players are dead, of whom s/t are scum on average and il est facile de voir that the average resulting state (s’,t’) = (s-s/t, t-2) has the same scum win rate when t is large. Treating s as a function of t, leads immediately to the differential equation
(2t+2) ds = s(t) dt
and thus asymptotically s = k sqrt(t)
Yeah, like I said, once I had a chance to check the actual numbers, it’s less extreme than I was remembering. I think the more extreme case I was thinking of might have come from mixing parities. For instance, with an odd number of players in the game, it takes 108 or 110 Townies to balance 5 Scum. I think I might have accidentally paired that number with even-number smaller cases when figuring the fitting function.
Run while you can! Starting mafia is just the gateway to addiction, which is the gateway to being burnt out, which is the gateway to quitting.
In all seriousness though, it provided me with many years of mind-engaging entertainment and community, for free even.
Newcomers are always welcome in the games I’ve played or modded. It’s fun to watch people who have no preconceived notions and who bring a fresh perspective.
I know there’s an “open” setup game getting started soon on a board I frequent (info here). An open setup is a good starting place because it means all of the roles are known in advance. Don’t get me wrong, the game is still confusing as hell, but that takes away one level of ambiguity.