[REVIEW] Open Setup Reviews

[Magua]

Re: Tit-for-Tat: Unless I'm missing something, this setups seems swingy in the wrong way -- in the case of the Mafia Jailkeeper or the Town Rolecop, dying not only punishes your side by losing a PR, but also helps the other side by giving them a PR.

I've only played the setup once (as scum) and didn't particularly enjoy that aspect.

[Magua]

I was busy!

Spoiler: Doing stuff...

I would switch the JK/Rolecop alignments, yes. The problem still remains that, normally, if you lynch the Mafia Rolecop, you're all "Yay, we lynched the Rolecop, go us!" but now you're all, "We lynched the Rolecop *and* got one on our side!"

That is, if the Mafia Rolecop is lynched D1, and the Mafia do not shoot the Deputy or the Jailkeeper N1, then they've pretty much lost. (This, I feel, is the same situation as now, if the Mafia Jailkeeper is lynched D1)

I think the overall setup might be more interesting as enablers -- that is:

-Town
6 Vanilla Townies
1 Rolecop-Enabler (if he dies, Rolecop can no longer rolecop)
1 Jailkeeper
1 Vigilante

- Mafia
1 Rolecop
1 Jailkeeper-Enabler (if he dies, Jailkeeper can no longer jailkeep)
1 Goon

But then that's a different setup altogether. There may already be a setup that's more like that. I dunno.

But back on track.

- I approve of switching Rolecop and Jailkeeper (and associated Backup) alignments.
- I don't support moving to 13p unless you trade the Vigilante in for a VT. That is, 10:3 is 4 mislynches to town loss (town picks 4 to die, mafia picks 3). With a vig, even one who constantly misvigs, town picks 5 to die, mafia picks 2.

[Magua]

^Wouldn't optimal strategy be for rolecop enabler to claim day 1 and vig to shoot him n1?

Maybe. I don't think I'd go with that strategy personally, but others might. It simply seems a waste to take a vig and use them to kill off what is essentially a confirmed town.

[Magua]

Not a big fan if scum in c9++ knowing exactly how much power town has by their roles. They also know if there is an sk, and the sk knows the setup. Part of the appeal on the old format was neither side knew exactly how much power was in the setup. At best a general idea.

Serial killer always knew the town power in the setup based on what powers they were given. That part is unchanged.

Mafia always knew within 1 the town power (2 goons + Godfather? 5T or 4T). The change here is that they probably know which of those it is, and whether there is a serial killer or not. Which isn't exactly ideal, but not a serious issue, IMO.

[Magua]

I see no reason why town would not give the scum the Encryptor each and every time. Rolecop essentially doubles the chance that scum will hit a PR rather than a VT after N1, which is pretty important in an open game.

I would replace it with 2-shot Janitor (scum learns roles kind).

For Redirector, it should be specified whether the tracker/cop know that they've been redirected -- eg, "Guilty" vs " Magua is guilty", or "Your target visited hoopla" vs "Magua visited hoopla"

[Magua]

I hate redirectors, so we're even. =P Maybe it's my formative experience, but I enjoy janitors in Open Games because they cut down on the necessity of mafia *having* to counterclaim a PR or die.

Other possibilities:
Roleblocker
Investigation Immune (both Tracker and Cop)
1-shot PGO

Ideally, of course, there should be no option that town will *always* take, and there should be no option that town would *never* take.

[Magua]

I believe the binning is because Assassin in the Palace's optimal strategy is "Bandwagon the very first vote in a Day, absolutely no talking"

[Magua]

No, of course it's not; it's actually worse to start a bad bandwagon in an AITP game then in a normal game. Just because some people don't understand how to play the game, doesn't make it a bad game.

I understand how to play the game just fine. I also understand how to play Dethy just fine. I've played fun games of Assassin in the Palace. I've played fun games of Dethy. Assassin in the Palace still has a breaking strategy that is not only unfun, but if followed gives the town a > 50% chance of winning based on EV, as does Dethy.

[Magua]

The first thing is that the "breaking stratagy" you're talking about ("lynch non-kings at random without giving scum any information") isn't actually possible. If you're lynching non-kings, then the scum will get information in the process. And you're likely to do worse then random, since the scum can also start wagons.

It is possible. And yes, it gives information -- the information it gives is that the person who was lynched is not the King. 'Cause they're dead. This is not useful information to the Assassin.

The Assassin can start a wagon, and this acts like a second shot for the Assassin -- if he starts a wagon on the King, the town loses, since they won't go through with the lynch (loss), but then the Assassin knows who the King is and can shoot him (loss). This has less of an effect that you might think -- an assassin-started D1 wagon in a 9 player game is only 1/8 to be successful, whereas an assassin-started D5 wagon is 1/4 to be successful -- but there's a 5/7 chance (or thereabouts) that the assassin will be dead before D5. The math balances out surprisingly well.

(This is, however, the reason that any given person is only allowed to start one wagon. Anyone trying to start a second before everyone alive has started one gets lynched.)

Second problem is that the more days the game goes before the assassin gets lynched, the worse the odds for the town get. Lynching at random is just a bad idea when you can do better, and you can do better.

The hypothesis isn't that the town can't do better than random lynching (hopefully, the town can; otherwise, we're all deluding ourselves here). The hypothesis is that the town has better than even odds simply *by* random lynching.

I'm not the math whiz that Llamarble is, so I fell back onto my strength: Monte Carlo simulations. You can download the code (executable is in the bin\debug directory) here. Source is included; it's C#.

It runs through however many games you want, with the following rules:

1) A random person who has not yet lynched is selected to be the lyncher. If everyone has lynched, that flag is reset and everyone can lynch again.
2) A random person is selected to be the lynchee. If the lyncher is not the Assassin, then the lynchee will not be the King.
3) If the King is lynched, Assassin wins.
4) If the Assassin is lynched, they take a shot at a random person that is not themselves. If it's the King, Assassin wins, if it's a Guard, town wins.
5) Otherwise, a Guard was lynched, and the game continues.

There is one other permutation, which I found interesting, which was the Assassin Lynch Day. This is the earliest Day that the Assassin will lynch, if possible (they may be forced to lynch earlier if, eg, everyone else has already lynched). If this is the blank, the Assassin will lynch whenever randomly selected (this is also equivalent to putting in 1). Otherwise, if it's before the Assassin Lynch Day, if they are randomly selected and are not the only viable candidate, someone else will be selected.

Running 1,000 permutations with 9 players gives these win % for the town:

Minimum Day Assassin Will Lynch	Town Win %
(None)	60%
1	60%
2	60%
3	61%
4	60%
5	60%
6	63%
7	64%

I'm not infallible, so feel free to examine the provided code, but barring glaring errors being pointed out to me, town wins ~60% just by random lynching. That's a broken game.

[Magua]

That's a shitty definition of "a broken game". A broken game is a game where a strategy exists that gives town better chances to win without scumhunting. You have no idea if town could improve on that 60% by scumhunting. I'm pretty sure they can, because if they scumhunt poorly and actually gain no information, you are back at that 60%.

Should find some time to talk to:

Just compare this with texas justice. That's a known broken game. Everyone shoots in a circle, and if scum isn't next to each other in that circle, town has won. And since the chances of winning that way are much better then with lynching and no strategy, it pays to not play the game and use the strategy. That's broken.

EV for the Texas Justice strategy is 73% (16 town vs 4 mafia). Could town do better than 73% with scumhunting? Maybe. Possibly. Town can certainly scumhunt and then try to pair people up to shoot rather than going by strict list ordering. Doesn't matter. Still broken. AitP is the same way.

EV for a Follow the Cop strategy in Basic 12 player can devolve to somewhere around 50% if the mafia claim/counterclaim the cop (and then shoot the cop if one of them isn't lynched), but that's still a breaking strategy.

Or, I guess, more generally I consider a breaking strategy something that is rote, formulaic, and still gives > 50% chance of success.

@Magna post 352: Simulation is fine to show what you're simulating - that town has a way to make their EV 60% - but doesn't address whether that is the town's best strategy, so you didn't show AitP to be broken. You did show that the balance needs adjustment. Would it be difficult to run the simulation again with a smaller game (1, 2, or 3 fewer guards)?

If it's not town's best strategy (and depending on the town, it may not be) that really just makes it worse for the Assassin.

The 60% was for 9-player AitP. 7 player random lynching is 53% win for the town. 6 player is 50%.

But keep in mind that that's the EV for random lynching, and, again, if you think the town can scumhunt better than randomly, you've pushed it townsided again.

I more like the idea that there are two assassins (with only one kill between them), because then that gives town something to talk about (possible connections between assassins). As it is now, I agree with IceGuy -- town really has no reason to discuss anything because town already has the information. You can say "But they don't know who the assassin is," but that just doesn't map. Yes, neither side has perfect information, but there's absolutely zero information that either side can give the other without losing them the game.

[Magua]

Oh, another thing about C9++ that I don't think we covered.

If a 1-shot role gets blocked, is their shot spent?

I say yes.

[Magua]

No. Scumhunting in a one-man scumgame is hard because there are no possible connections to find. Everyone has a vested interest in not being the person lynched (since it's bad for their team, no matter who they are), there's no opposing team that makes scum worry about not being "too effective" and being Night-killed, and there's no other people the one-man scumteam has an interest in not dying that can be used as connections.

[Magua]

Been a long time since I've been in this thread. =P

Mmmmmm.

Setup just seems annoying.

- Setup is townsided if the scum join forces while the jailkeeper is still alive (compare to say 10v2 Mountainous, this is 11v2 w/ Jailkeeper).
- Setup is *probably* townsided while the scum are separate. Not only do you have the Cop, but the Jailkeeper can be functionally like a cop as well, and if the town lynches scum, see above.
- Setup allows for practically zero scumhunting, certainly zero connections until there's a join.
- If the mafia crosskill, it's almost certainly town win.

I'm not sure what I would recommend changing -- I tend to like my mafia's informed from the beginning and this is about the opposite.