[SETUP] Shady Cops

[BNL]

I might have submitted this for the January challenge if I had thought of it earlier, but alas

3 Mafia Goon
8 Town Cop-Miller

Whenever a Cop investigates, it permanently appears guilty to other Cops, for that Night and all future Nights

[Awoo]

So if two people claim an innocent result on the same person that's a 50/50 chance to find a mafia, with the second person to claim being more believable.

Mafia can safely inno each other, but too much incest might be suspicious after one flips.

[Creature]

Still feel this can be broken if players are divided into two groups with one investigating the other.

[callforjudgement]

Still feel this can be broken if players are divided into two groups with one investigating the other.

I agree. I can't see an outright break on N1 (there might be one, though), but the game continues beyond that and it should be possible to solve the resulting logic puzzle. There are a huge number of cases to consider, though.

My initial thoughts were to split the 10 remaining players 5 and 5 N1, and have all the actions pre-arranged to prevent collisions, but after more thought it's probably better for the investigating 5 (the 5 more townish players) to select their targets from among the non-investigators secretly and purely at random (using nonprovable randomness); it's not as though collisions are useless, after all. Scum can't break up the strategy with their nightkill because the targets are secret and (being random) there's no way to predict what they are psychologically; the randomness also helps to avoid everyone piling onto the scummiest or nulliest player. D2 the investigating millers claim results; fakeclaiming results as scum is very risky unless they're near the end of the claim order, because there's a considerable chance of a duplicate. (If you do get two conflicting claimed results, you lynch the player who claimed their target was scum; if that player was telling the truth, you've caught

two

scum, which more than makes up for the mislynched townie, and if they were lying, you just lynched scum.) As a result, the first couple of results to be claimed are likely highly trustworthy; if someone's claimed to have been investigated as scum in the first two results, you lynch them, and if scum killed in the pool of investigators the first player who's claimed to have been investigated as town gets moved into the investigator pool to replace them. You continue "investigate a random player from the pool" strategies on future nights too, although you remove people you've already investigated from the pool you're randomizing from. Investigating players have to be sorted via dayplay or the scum nightkill, but the night actions nonetheless give a huge boost to town.

Calculating an EV for this would require knowing the best scum counterstrategy (in particular, when should scum inside the investigator pool lie about their results?), so it'll be quite complex to figure out whether this is a true break or not unless someone finds a counter-counterstrategy for town that works in all cases.

Another possibility I thought of was to split the players 4 and 6 (and another is to not replace dead investigators); I suspect this is worse than the 5:5 split with replacement, but am not sure (this is mostly just a guess).