Open 28 - Quack Mafia. OVER! But who won? before 466

[mneme]

SSF: I was just going to say: "Hi, I'm here -- will post more once I've read the thread!"

Oddly enough...

[mneme]

Alas -- Roach missed his choice. The perils of replacing.

[mneme]

Ok, I've read through. People on the d3isted lynch (not sure how much this matters):

JordanA24, Stewie, curiouskarmadog, mneme, HonaryHitchhiker, somestrangeflea

The d3isted lynch is interesting for a number of reasons.

First, it's pretty obvious (clearly, my predecessors didn't hink this way) that d3isted shouldn't have been lynched on most grounds. His initial claim of "I know I'm a Quack" was valid and one of the best claims we had out there. Moreover, despite being dismissed by all and sundry, his claim that the strongest strategy was for all Doctors to chose "no target" (thus reducing the game to a roles-revealed-on-death Vanilla game with any given townie having a 1/2 townkill/save ability for the endgame)) was sound, and was never really discussed in the depth it deserved --it is, however, too late for such radical theories to prevail, since we've reached endgameland (unless we want to no-lynch today, no-target tonight, and hope to lynch 3 scum in a row). It's possible that it was also suboptimal (I certainly haven't done the math on the current game -- but I'll note that with a very low number of samples to perform calculus on before reaching endgame, and few unknowns filled in on death, there are fewer positives to "chose every night" than it seems at first, and the negatives are obvious).

That said, I think he was lynched more because of craplogic (the garbage about wiping the slate clean between replacements, frex), even if that did obscure some good points.

That said...looking at the results from last night, I see (not sure there's a lot here, but might as well express sylogisms anyway):

Hjallti died (and was Town), and was targetted by only Tarhalindur. Possiblities: Tar is scum. Tar is a quack. Tar is a doc and Vel-Rahn Koon (as quack) and the scum also targetted Hjalti (too low to count).

Jdodge lived and was targetted by curiouskarmadog and somestrangeflea. Possiblities: Jdodge is scum. At least one of ckd and ssf are docs. ckd and ssf are both scum.

Jordan lived and was targetted by Stewie. Possibliiteis; Stewie is a doc. Jordan is scum. Stewie is scum. Both are scum.

Vel-Rahn Koon died and was targetted only by Jordan. Possiblities; Jordin is scum. Jordin is a quack.

ssf lived and was targetted by Jdodge and HH. Possiblities: see Jdodge.

[mneme]

Stewie: there's a bit more to it than that (what we have are a set of X equasions that can be solved together) but at best, they give us a set of unknowns that are very difficult to fill in; I find it unlikely that we can solve the entire set sufficiently to be able to prove innocent (or guilty) very many people. But we may be able to get odds if we crunch the numbers enough.

Worth it? Not sure.

[mneme]

Jordan: It's a lot of work to figure out who is/might be scum based on claims, and I'm not sure I can get the math right. So while I'm interested in the work being done, I'm unconfident enough in my ability to get it right not to try it right now.

Now, in terms of analyzing day posts, that's easier (if less sure), but IMO it's a lot less fruitlful than trying to solve the night calculus.

[mneme]

As I mentioned (iirc) in an earlier post, this is a sucker bet game -- quacks have eliminated more town than scum have (obviously). We'd probably have much better odds had everyone either agreed to protect nobody, or to randomly choose between one of two people, or otherwise actually played more systematically.

So.. JDodge's claims about "no lynch" were entirely clear; we're in endgame with even numbers, so if we choose to go vanilla or mostly vanilla (with anyone who can't be close to absolutely sure they are a doctor -- and that's -probably- everyone, factoring in dead-person protects and scum kills, but someone might be able to get a high probablity of self-doctorness, since they can rule out the "or the targetter is lying" cases we have to include --refraining from a -- refraining from protecting), "No Lynch" is correct.

If one assumes vanilla play, the probability of the town winning without a no-lynch is 3/8*2/6*1/4 = 1/32.

If we no-lynch, the scum will probably kill a townie for us (and if that stalemates, we can start doing night gambits with maybe-doctors which will probably either give us enough info to be useful or reduce us to the good numbers, so it's all good), and we can play vanilla to try to lynch three times in a row...at 3/7*2/5*1/3 = 2/35. In other words, we almost double our chances of winning if we no-lynch today.

Also, protecting scummy players is correct; the town doesn't win by figuring out the setup. we win by not killing too many townies and figuring out who the scum are.

[mneme]

(protecting scummy players if you think you're a likely quack).

[mneme]

cmk: I was interested in formulating a strategy, since it looks to me as if we're doing a great job killing ourselves, and despite JDodge's inherent scumminess, it's possible there was a point there.

But...lynch all liars is an -important- meme, and Dodge is clearly scum.

unvote
vote: JDodge

[mneme]

With the way the quacks work, I'm not sure the town -isn't- better off just playing as newb; the reward of getting a bunch of doctors cross-protecting, but it does seem like the town's going to lose a bunch of people doing it. The theory I was propounding on day 2 was sound (if you don't -know- you're a doctor, don't target; question of odds of lynch vs no-lynch and lylo next day); note that the Mafia, as it turned out, could have missed our kill again and still won (because while Jordan had figured out he was a Quack, Tar hadn't).

The setup -was- balanced; a vanilla with 9 town to 3 mafia is pretty balanced, and this is better than that, since the town does have powers. But I'm not convinced it's balanced with the strategy actually taken (ie, protect every night until you find out your flavor), as the reward isn't quite worth the gain, by my analysis. If flavor was revealed on death, it would be, I think, since you'd know how many dead quacks/docs there were, but without that, you don't really get to catch scum with all the kills (at best, you narrow things down a bit), and you do the mafia's work for them.

[mneme]

Tar: er. If you'd figured out you were a quack, why target? To act as a "kill cop?" The problem is, the town was at single death loss numbers, so that was perhaps not ideal; if you were wrong, the town lost (as it was, we lost anyway, but that wasn't true if we didn't lure the real doc protects away from the town).

Good Game! (even if I did replace into it late).

[mneme]

Hmm. Ignoring the vanilla strategy, I think the town may benefit most from -only- sharing info on dead people -- because this avoids just -handing- docs to the scum. Particularly since scum are much more likely to fake doc than quack regardless. If someone died, sharing info on protects on that person -only- may give you a good idea of who might be a quack (frex, if two people protected a dead person, chances are that either scum targetted that person or they're both quacks... And most often, only one person will have targetted the dead, and that person is a quack or scum, and now knows it.

Trying to use quacks as cops is a possible strategy in some situations, but is -usually- a bad idea; the town benefits most under most situations by known quacks just stopping targetting people.

Now, whether this total (always target until you're pretty sure of your flavor, announce only successful "kills", stop kiling once you know you're a quack) is better than vanilla is an open question, but I think it may be better than the town's strategy this game.