Matrix12 setup - Expanding the Matrix6 setup

[TheMainMan]

Hi guys

I have tried to expand the Matrix6 setup so that it features more players (17) and a few additional roles.

What are your thoughts on the balance of this new setup?



For those wondering, some of the roles in the matrix (Mafia Ninja and 1-Shot Mafia Bulletproof) will indeed only have value for one of the rows or columns setup in which it is used and not the other. This is not a mistake, it is done on purpose to only boost the mafia side in one of the two setups.

The Innocent Child may choose himself when he wishes to be publicly confirmed in the thread as town in this setup, rather than it being announced automatically at the start of the game.

The Even Night Vigilante may only shoot on even nights (i.e. Night 2, Night 4, etc.).

[Accountant]

I don't know, since you didn't tell us the setup.

[Alchemist21]

Does 1 save for the Mafia JOAT mean 1-shot Doc protection?

[TheMainMan]

I don't know, since you didn't tell us the setup.

What info are you missing?

The host randomizes a letter or a number, deciding which setup is used. I.e. if the host rands C, then the setup will consist of the roles in the C column. The host then adds 2 Mafia Goons to the setup and fills up the rest with Vanilla Townies.

I.e. setup C would give you this:

1x Town Doctor
1x Town Tracker
1x Town Vigilante
10x Vanilla Townie

1x Mafia Role Cop
1x Mafia Jailkeeper
2x Mafia Goon

Does 1 save for the Mafia JOAT mean 1-shot Doc protection?

Correct.

[Accountant]

Huh. Image didn't show up on my browser. Would this be used in newbie games?

[Davsto]

I assume this is merely meant as a separate semi-open setup to matrix6 and not to replace it as the current newbie setup

I'm not great at balance so I won't comment but how do the Masons work in this?

[TheMainMan]

Huh. Image didn't show up on my browser. Would this be used in newbie games?

I assume this is merely meant as a separate semi-open setup to matrix6 and not to replace it as the current newbie setup

I'm not great at balance so I won't comment but how do the Masons work in this?

Yeah, what Davsto said.

The Masons are two townies who will have it confirmed to them by the mod that they are both townies and can trust each other, and they will additionally have 24/7 private chat capabilities, i.e. they will have a private thread where they can always talk with each other.

[callforjudgement]

One potential problem is that unless I missed a combination, scum always know which setup they're in.

Although I'm not an expert at balancing games larger than 13p, I checked a few setups at random and they all

seemed

scumsided. Setup 6 is innocent child + tracker + 11 versus roleblocker + 3; I have a hard time believing that this isn't scumsided. (I decided to work out what a typical D1 and D2 that leaned slightly town would look like, and came up with "VT and Goon lynched, two VTs killed, tracker gets "visited nobody" both times". This would reduce the setup to innocent child + tracker + 8 versus roleblocker + 2, which is highly scumsided even as an Open; town would have a few associatives more than usual from the scumflip, but I don't think that compensates.)

As another example, take setup B: we can assume that the Cop trades themself for one scum (what Cop wouldn't claim when they got a guilty?) Town have no other power roles, so if the Cop finds scum N0, this is equivalent to a 12:3 mountainous. If town now lynch scum the next day, it becomes 11:2 mountainous, a setup so scumsided that AFAIK town have never won it on mafiascum.net. Getting a Cop guilty + a lynch on scum is what would normally be considered the

best

case for town, and here it is leaving town with basically no chance. (What they need to do in the setup is to keep the Cop alive as long as possible and hope they can claim before they flip, but they've been given no method via which to accomplish this.)

I don't think the idea is terrible, but the balance seems to be all wrong.

[Davsto]

One potential problem is that unless I missed a combination, scum always know which setup they're in.

This. The only two setups in which scum doesn't know which setup they're in is A and E, and that's just a 1/2 chance of figuring it out, and with an Innocent Child in the first and n0 Cop in the second, it's not gonna be hard for them to figure out.

[TheMainMan]

One potential problem is that unless I missed a combination, scum always know which setup they're in.

Although I'm not an expert at balancing games larger than 13p, I checked a few setups at random and they all

seemed

scumsided. Setup 6 is innocent child + tracker + 11 versus roleblocker + 3; I have a hard time believing that this isn't scumsided. (I decided to work out what a typical D1 and D2 that leaned slightly town would look like, and came up with "VT and Goon lynched, two VTs killed, tracker gets "visited nobody" both times". This would reduce the setup to innocent child + tracker + 8 versus roleblocker + 2, which is highly scumsided even as an Open; town would have a few associatives more than usual from the scumflip, but I don't think that compensates.)

As another example, take setup B: we can assume that the Cop trades themself for one scum (what Cop wouldn't claim when they got a guilty?) Town have no other power roles, so if the Cop finds scum N0, this is equivalent to a 12:3 mountainous. If town now lynch scum the next day, it becomes 11:2 mountainous, a setup so scumsided that AFAIK town have never won it on mafiascum.net. Getting a Cop guilty + a lynch on scum is what would normally be considered the

best

case for town, and here it is leaving town with basically no chance. (What they need to do in the setup is to keep the Cop alive as long as possible and hope they can claim before they flip, but they've been given no method via which to accomplish this.)

I don't think the idea is terrible, but the balance seems to be all wrong.

The scum knowing what setup they're in is just part of the setup and the balance for this setup. I know it's not like that in Matrix6, but it was difficult to prevent here, so just had to go with it. If you feel town should be boosted in some of the setups because of it, I'm very open to suggestions.

Re: Setup B that is actually a tried and true setup that's been run 100+ times on another site, and the town wins 45% of the time, so it should be at least close to balanced. It's also basically just the same as one of the setups in Matrix6, just scaled for more players.

I don't know how I feel about the 1-shot mafia vigilantes and such, so if you have any suggestions for other powerful mafia roles to put in the setup, that might help improve the setup.

[zMuffinMan]

What are your thoughts on the balance of this new setup?

it seems virtually impossible to balance something like that while maintaining a decent variety

if balance is your primary concern, don't run a setup like that

also, without running any actual numbers on any of those, i'd hazard a guess that the vast majority, if not all of those setups, are scum-sided (not including swing, and some of those setups look like they have potential for very hard snowballing)

I don't know how I feel about the 1-shot mafia vigilantes and such, so if you have any suggestions for other powerful mafia roles to put in the setup, that might help improve the setup.

um... why does scum need powerful roles in a 13:4? 13:4 is inherently scum-sided by numbers alone unless town is somewhat power-heavy... honestly, i think you could run most of those setups with 4 goons and scum would probably still have an advantage (i may actually write something to check whether or not this is true)

[Accountant]

Balancing a matrix17 means you have to check for balance on 12 different lists and make sure they intersect in a neat matrix.

[callforjudgement]

Re: Setup B that is actually a tried and true setup that's been run 100+ times on another site, and the town wins 45% of the time, so it should be at least close to balanced.

Either they have a very townsided meta, or we have a very scumsided meta. Probably both.

Just to be clear, we're talking about Cop + 12 versus 4 with Cop Head Start. I can't see any plausible way for town to win this if the Cop is exposed early (for

any

reason, including finding a guilty), so they're going to have to survive as long as possible without claiming (ideally leaving breadcrumbs that nobody picks up on until they're dead), and hopefully build up a number of investigations in that time that allow for forming a townblock or eliminating multiple scum. I guess I can see town winning if the Cop is very good at not dying, but the odds of that are pretty low as you have to dodge both the lynch and the nightkill. (I can imagine a hypothetical site meta in which lurkers are left alone by both town and scum until close to deadline; that would probably allow the setup in question to be close to balanced, but IMO if that were the usual way to play, scum would be playing suboptimally.)

[Accountant]

A cop is nowhere near powerful enough to overcome the hilariously overpowered 4 scum vs 13 town setup

[TheMainMan]

It is indeed interesting how balance is perceived differently on different sites.

The site where those stats are pulled from is a very active site that will often get 1000+ posts for the first few day phases of a game that size, and the day phases last 12 hours. So all players are active and it's often an environment consisting of real-time interaction, which can maybe help out the town getting better reads. There's also a meta of covering for the cop, hypocop'ing, where each player will leave a "fake cop check" each day as if they were the cop. That means that when someone dies, you can often clear some people based on who the dead person fake cop checked (this is ofc sometimes exploited by the mafia).

On that site, the stats clearly shows that that setup is not hilariously scumsided.

[callforjudgement]

Out of interest, what are the stats for when the Cop normally dies in that setup on that site (in terms of day/night phase)? Even better if you have a correlation between that and the result of the game.

[PantherPunt]

Ya I don't think the "leaving cop cover" play is understood here as those setups don't typically run here. Like a common setup is:

9 players:
2 goons
1 cop
6 vt

the vt's (and often the goons) leave fake investigations in addition to the real investigation left by the cop; i.e. everyone posts "If I am cop I investigated Player X as town"

Now, the incorrect fakes make it easier for the mafia to POE the cop, but assuming multiple people leave "correct" fake investigations, then it can often result in the cop living multiple days while leaving their investigations in the thread with no ambiguity.

There is also the concept of "cop hunt cleared" where, for instance, Player A leaves an investigation of "If I am cop I investigated Player B as town." If Player A is nightkilled, it's reasonable to assume that the investigation was a "correct" fake and that Player B is town. As mentioned, this can be exploited, but it means that the mafia are intentionally leaving the cop alive for another night, which isn't great EV.

[callforjudgement]

Right, I was beginning to think along those lines. (And if nothing else, it's a demonstration of just how townsided the mere fact of a setup being Open can be!) I'm kind-of reminded of Low Effort Mafia, which has similar properties and is also heavily townsided. Additionally, PR hunting is quite difficult if every VT knows that they're trying to protect one power role in particular. I'm curious enough to run a simulation now, just trying to work out what the assumptions should be.

This actually makes me even more convinced that the setup is imbalanced in Matrix12, because scum know they're in the setup but town (initially) don't, and thus town quite possibly don't have the evidence to know they need to take actions to protect a Cop.

[callforjudgement]

OK, so here's my simulation results for Setup B as an Open:

Code: Select all

Town wins: 2639; Scum wins: 7361
When the cop dies D1:  town wins:    81; scum wins:   500
When the cop dies D2:  town wins:   107; scum wins:   476
When the cop dies D3:  town wins:   125; scum wins:   499
When the cop dies D4:  town wins:   136; scum wins:   450
When the cop dies D5:  town wins:   170; scum wins:   378
When the cop dies D6:  town wins:   155; scum wins:   304
When the cop dies D7:  town wins:    85; scum wins:   239
When the cop dies D8:  town wins:     0; scum wins:   185
When the cop survives: town wins:   367; scum wins:  1603
When the cop dies N1:  town wins:   109; scum wins:   596
When the cop dies N2:  town wins:   136; scum wins:   597
When the cop dies N3:  town wins:   176; scum wins:   550
When the cop dies N4:  town wins:   227; scum wins:   488
When the cop dies N5:  town wins:   255; scum wins:   351
When the cop dies N6:  town wins:   299; scum wins:   133
When the cop dies N7:  town wins:   211; scum wins:    12

This is assuming that the town lynches regardless of what the lynchee claims, that vanilla townies and scum both hypoclaim on a random living player they haven't investigated yet (other than themselves) and claim a guilty the same proportion of the time as a Cop would be expected to scan a guilty, and that scum aim to kill players who have had a lot of "innocent" hypoclaims on them or innocents from a flipped Cop (as opposed to trying to kill the Cop, which would allow cop hunt clearing). It also assumes that town random-lynches until the Cop dies, then lynches all guilties in sequence, then lynches among the pool of unconfirmed players.

This isn't perfect. Factors that make the result more scumsided than it should be include the probability that the Cop will claim at lylo; although scum will counterclaim (in fact, quite possibly multiple scum will counterclaim each other and the real Cop), this reduces the size of the lynch pool a little. Factors that make the result more townsided than it should be include the possibility that scum can seek to kill the Cop early without much risk of a cop hunt clear (by doing it with less than 100% probability), and that scum can intentionally refuse to kill probable Cops if they survive into the lategame in order to cast doubt on their results.

The ~26% town EV for Setup B is about where I was expecting it ro be, and IMO too low to produce a balanced game (the usual EV for a balanced game is around 40%; typically people use values less than 50% because town can catch scum via dayplay too). If scum are losing half the time, they are probably making theory mistakes (I suspect they're overvaluing the value of a Cop death and leaking information to the town while trying to get it).

I should note that the 9p version (Cop + 6 versus 2) is run on Mafiascum.net frequently (in fact, it's a possible Newbie setup) and believed to be balanced as part of a semi-open (it's been in a bunch of semi-open Newbie setups over the years), although that's without an N0 investigation (if you add the investigation it becomes townsided). It's the 17p version that I'm taking issues with.

[TheMainMan]

Thank you for that post. It's definitely enlightening and I've seen similar simulations before. They usually don't give a clear or correct picture though, judging from experience and large sample sizes of actual games.

Nonetheless, I do agree that setup B suffers from the fact that the town doesn't even know if there's a cop. I've added an Innocent Child to the setup to combat that and have also made some other changes to nerf the scum a bit in most of the setups:

[Davsto]

scum always know which setup they're in.

[TheMainMan]

Yeah, I know. Like I said earlier, that's just part of this setup. I completely agree it changes the balance, but then that just has to be accounted for. If you think it isn't accounted for and town needs further boosting or scum needs further nerfing in any of the setups, please let me know.

[PantherPunt]

Are there by chance approved normal 17er setups that can be crossed up to formulate a grid?

Or at least a few that can serve as the basis for a grid that would prevent scum from knowing the setup?

[Ircher]

Mafia Vigs are bastard.

[PantherPunt]

Lolwut