[OLD] Open Setup Ideas and Discussion

[callforjudgement]

I think there's a temptation to argue that with optimal play, town can do better than random, but this neglects the fact that EV assumes

both

sides are playing optimally.

That is indeed a good point. I guess the counterpoint is "if town think they're going to play below EV, they can just agree to random-lynch" (which is where IIRC the commonly seen rule against provable randomness comes form), so if you assume that the players are perfect

strategists

, but not necessarily perfect

scumhunters

, then you'd expect town to play at EV or better. (Or in other words, the argument is just a disconnect about what "optimal play" means. You're working from the assumption that players are perfect at strategy, perfect at hiding as scum, and perfect at scumhunting. I think the more commonly seen assumption in EV theory is that players are perfect at strategy but have no idea how to scumhunt, and thus it doesn't matter how good the scum are at hiding. The two assumptions give the same EV, and thus it normally doesn't matter which set of assumptions you're talking about, but plausible small deviations from the first scenario leave town's win rate untouched, whereas plausible small deviations from the second scenario increase it.)

In practice, of course, players often suck at strategy in addition to sucking at scumhunting, and scum have counterplay in practice that can make town's strategy worse (for example, many players (correctly or not) believe theory discussion to be a scumtell, and thus you can sometimes argue towns out of agreeing to follow a breaking strategy). It's rare for towns to random-lynch even when that would be their best option, which is one common reason for towns to play below EV.

[callforjudgement]

Seriously, if you want to make this setup work, you have to go the other way. Something like 6 VT, 2 Mafia Lynchers, nightless, scum win if

both

their lynch targets are lynched has an EV of slightly below 50% (which is where you want it) and isn't breakable. (Lynch targets should be unaware but should probably

flip

as lyncher targets, so that the town knows how well they're doing.)

I'd be pretty curious as to what the win rates in practice would look like in this setup.

[callforjudgement]

Nominate three players to shoot, and give them each a target who's outside the set of three shooters.

If any of the targets survive, the shooter is confirmed scum. For any target who dies, the shooter is confirmed town. So even if all three targets die, you have five players, of which three are confirmed. Three confirmed townies and two confirmed scum are both an instant town win. The remaining situation has 2 confirmed townies, 1 confirmed scum, and three unconfirmed players: lynch the scum, and get the unconfirmed players (who each have their shot left if town) to vig each other in a cycle for the win.

100% town EV.

[callforjudgement]

Who chooses which player the investigative result is about?

Anyway, I suspect this is townsided, because the Copmakers can prove their towniness via claiming (or else end up in a 1v1); they don't have to fear being NKed by scum as that doesn't actually hurt town's chances, so they can just claim as soon as they get the role. It has a lot of similarities with the closed/semiopen setup Cop + 6 VT vs. 2 Goon (which is known to be balanced); both have one investigation result per night, and one player who can claim and is either counterclaimed by scum (leading to a 1v1) or confirmed. The difference is that there's no way for the Cop to die early here (they can only be suppressed for one night), so some of the worse possibilities for town are much less likely, and that probably makes the setup townsided overall.

[callforjudgement]

You could have the setup vote done via PM, then have the votes revealed at the end of the signups. At that point, scum would know that they were scum and could fakeclaim with knowledge of that. (This would require a plurality rather than majority vote.)

I suspect it'd be townsided, though, as there's still a lot of things that could trip scum up. In particular, do they fakeclaim what their roles in the nonchosen setups were too?

[callforjudgement]

As I see it, the basic problem is that town gets more benefit from spamming their roles than scum do. One cop scan gives you one confirmation. Two vig kills gives you two confirmations (because if you shoot town twice, you end up gaining a mislynch which neatly cancels out the two dead townies). The JK probably shouldn't act until later in the game, because JK shots are more valuable the longer the game goes on and you want to conserve roles for the other power roles early. Meanwhile, the scum's power roles are a lot more marginal. Framer doesn't do anything unless it hits the cop's target (and scum's frame is only one-way; inno results from the cop are reliable, and more common than guilty results). Doctor doesn't do anything unless it hits the Vigilante's target. Rolecop is the most useful role there, and it doesn't have an immediate effect (basically it lets you shut down a power role two nights later, because it takes one night to scan them and another to NK them).

In a shared-shot-reservoir setup like this, therefore, having a larger number of shots available will help out town much more than it will scum. Scum don't really have much to do with their number of shots (and in Open 634 which you linked, it's possible to observe scum wasting their shots on marginal actions simply because they had no better use). Town, meanwhile, can make excellent use of all six of their shots. That is, I suspect, what causes the townsidedness of the setup.

In order to fix it, either the scum need more useful roles, or the number of total shots available needs to start lower. JK versus JK sounds like an action resolution nightmare to me, so might not be the best fix, but it's thinking along the right lines. I might also consider decreasing X and Y to 3. (The setup is obviously scumsided with X and Y at 0, so there's likely a balanced number somewhere in the range.)

[callforjudgement]

Arguably, the only balanced EV in multiball is 50/50/50, but that'd take some very bizarre win conditions to get right.

Speaking more seriously, it's unclear to me whether 50/25/25 or 33/33/33 is a better balance to aim for in a multiball game.

[callforjudgement]

This is sort of separate but
Are there any matrix6 games that are not in newb queue?

Micro 180, Micro 188, Micro 247. (Only one's linked, because I found the other two by searching for the signup threads.)

[callforjudgement]

I think it might be worthwhile to have a thread (either this one or one in Open Setups) that people could throw out quick experimental setup ideas in to see if they got shot down, but a forum sounds like a good idea for more "serious" setup ideas. (We could/should probably also do Open Setup reviews in the setup thread rather than in the Open Setup Reviews thread, too, although then we might need some other method to give them visibility. The change is nonetheless a good idea, though, especially as people often miss the purpose of Open Setup Reviews and confuse it with this thread.)

[callforjudgement]

That's because most posts about Opens should be in this thread.

(The other one's for setups that are about to be run, to get input from the Open listmod and/or other players about whether it's balanced enough to run.)

[callforjudgement]

Delayed Kill

Interesting mechanic. Mathematically the mechanic's equivalent to giving town a factional doublevote (because you're basically just giving a dead townie a vote), which helps the town because they gain an extra mislynch at the end of the game (when scum are easiest to find via PoE). As such, it's notably more valuable than an extra VT. That said, 2:8 nightstart is highly scumsided, so I suspect this game is scumsided too even though it's more townsided than that.

EV calculations for the mechanic (given are the numbers going into night); I did these in my head (writing down working at intermediate stages when I got confused) so they might be wrong:

2:1 1/2
4:1 5/8	3:2 1/4
6:1 33/48	5:2 3/8	4:3 1/8
8:1 279/384	7:2 29/64	6:3 7/32	5:4 1/16
	9:2 195/384	8:3 37/128	7:4 1/8	6:5 1/32

Assuming my maths is correct, 29/64 is not that far off where we'd want the EV of a game to be, so 7:2 at least seems reasonable. However, given that the game is effectively mountainous and scum have a full nightkill every night, town tends to do considerably worse than EV, so I'd probably throw in some extra VTs; if I were running this setup, I'd probably start at 9:2 (which has an EV of just marginally over 50%).

[callforjudgement]

This brings up the right opportunity to ask this (I think) with 2:8, 6 is required for lynch, which means more Townies need to be on a Scum lynch (if Scum are not bussing). How does/does this have an impact on things at all?

EV theory assumes that this will be counteracted via the tendency of scum to bus (because if scum skewed their votes towards not bussing, town could use that as a method of determining who the scum were, bringing up town's win rate a different way). This probably doesn't exactly balance in practice, but it's good enough for approximations.

Instead, we allow for these sorts of things by comparing EVs to observed win rates. For example, in a mountainous game in which scum have a nightkill, observed win rates are much lower than EVs. The most likely explanation is that, with no theory-based reason to kill one townie or another, scum just kill all the best scumhunters early in the game.