Newbie Setup (Matrix6 implemented)

[callforjudgement]

The EpicMafia stock "newbie" setup is probably a bad idea here. It's balanced 50-50… if you follow a breaking strategy that leaves most players with not much say in the game. (You basically always get two cop claims day 1, one from town, one from scum, unless the cop died overnight.)

Knowing a game's setup has a significant effect on the win rate (this isn't exact, but approximately, it makes all town power roles into Innocent Children because it's normally a bad idea for the scum to counterclaim them). Open setups are balanced knowing that the player will know the setup, so if they're run as closed, the setups have to be more townsided. And I think that semi-open setups are probably better than opens for newbies; it's going to be better preparation for the rest of the site.

Btw, my own experience with cops in 2of4 is that scum are basically screwed unless they get rid of them by or on night 2 (but also, that they have a good chance of getting rid of them by or on night 2, which is what keeps the win rates down). If the cop makes it to day 3, scum basically have to argue that the cop is actually scum, regardless of any evidence to the contrary. (Better hope they're both still alive!)

The proposal seems to be for a semi-open with a great number of possible setups, and the players don't determine which is used until the game is over. I think this is reasonable (probably around 7 or 8 setups is the right number). One interesting alternative might be to go fully closed, but that would require some sort of Newbie Review Group to maintain a secret list of balanced 9p closed setups and add to it from time to time, which is likely a lot of work. (The advantage of this over semi-open is that it avoids requiring a huge overlap of setups to avoid the phenomenon of "well, you're a cop, so we know there isn't both a tracker and a rolecop in the game", or the like; semi-open with a large stable makes fakeclaiming very confusing.)

Incidentally, I proposed a potential-for-newbie setup here, but nobody commented on it. Perhaps I'll run it over in the Micro queue when my current game ends. It's probably simpler to explain than 2of4, too, although it leads to quite a long list of "all possible role PMs" in the second post. This approach of a single semi-open with lots of possibilities that follow the same rules may be better than drawing from a stable of radically different setups to avoid weird breaking strategies involving the interplay between different setups. (You don't want town knowing they're in a cop/doc setup because of a roleblocker flip, if the setup isn't designed for that…)

[callforjudgement]

The reason to have a balanced setup is that playing in an unbalanced setup isn't as fun (especially if it's won by the side it's unbalanced in favour of, which mathematically most of the time it will be).

[callforjudgement]

One thing I wonder about is why the SE queue is so mindboggingly long. Is it people trying to play their third game in a setup they're familiar with by now? Or experienced players who just feel like having a nice small game with some inexperienced players?

One potential fix might be to add one SE and remove one newbie from each newbie game. This seems to be better for the newbies (they get their games started faster, and more experienced players to talk to when the ICs inevitably die night 1), and better for the SEs. The only potential problem I see is that, with more newbie games starting, you might run out of ICs or mods faster.

[callforjudgement]

Btw, I review quite a few setups (I've basically been camping the review queue for 9p reviews), and considering the worst possible scenario for town is one of the things you do. In most reasonably balanced setups, the town's worst case scenario day 2 is 2:5 mountainous. If it's better than that, you have the issue that scum can't really gain much of an advantage no matter how well they play. (A typical best case scenario for town on day 2 is one scum dead and a decent lead on the other. That's probably even more powerful in town's favour than 2:5 is in scum's. Although, the worst case scenario is typically a bit more likely than the best case scenario.)

[callforjudgement]

I'd actually say the major issue at 9p is swing. Getting the setup to within the vicinity of balance isn't too hard, but you're sorely constrained in what you can do without making the game likely to become townsided or scumsided pretty early. (As an example, even Cop + 6 VT + 2 Goon, which is a reasonably balanced setup, is pretty swingy; if the cop survives to day 3, town still have a pretty high chance of winning even despite having mislynched twice, and if they die night 1, it's mountainous.)

[callforjudgement]

Yeah, JK's even worse than Cop in the swing stakes.

(Doctor is the least swingy role commonly seen in newbie games, but it's also easily the least powerful.)

[callforjudgement]

So you'd kill any sort of connection hunting in order to teach scumhunting better?

Also, that setup is scumsided. The Cop doesn't really do anything useful (except confirm a small subset of the townies if the Godfather dies), so you have Doctor + 6 VT against two Goon, and that setup is definitely scumsided. And on top of that, you have unreliable flips!

[callforjudgement]

Fastest way: go to an old newbie signup thread, look at the announced playerlists, look at those users' posts to see if they still exist.

[callforjudgement]

Would the players know which setup's been selected?

If yes, you probably have the same issues as you do at the moment.
If no, you have the issue of having to be an expert on all of the possible setups to get the theory right.

[callforjudgement]

Yep, basically. You end up with crazy setup breaking like "well, there's a doc in the setup, so there can't be both tracker and JK…", meaning that in general it's very hard for scum to figure out what they can safely fakeclaim.

[callforjudgement]

The problem's basically that it gives you a reward for obscure setup information newbies couldn't really be possibly expected to know. So it's basically putting them at a disadvantage based on things that shouldn't be relevant.

The claim-busting thing you suggest is fine, but if it's happening

only

to newbies, we have a sort of discrimination problem there.

[callforjudgement]

That's because the power of a Doctor relates a lot to what they have to protect.

If the best a Doctor can do is to protect particularly townish-looking VTs, it's not going to help the town that much; they're going to have to land the protection twice before town gets much advantage from it, so it's not a huge discouragement to scum shooting the town's top townreads in order to gain an advantage; if the doctor hits for the first time, they're going to have to randomize their target a bit more the second time, but this isn't a huge detriment to scum. Not as bad as mountainous, but not nearly powerful enough to be balanced in an otherwise mountainous 2:7.

If Doctors can protect power roles, on the other hand (particularly investigative roles like Cop and Jailkeeper), they become a lot more useful.

[callforjudgement]

Did the makeup of the players in the game change, too? IIRC there used to be four ICs, rather than just one, but I don't know whether that change was made before or after the first set of statistics.

[callforjudgement]

Who's the backup list mod for the Newbie queue?

This sounds like the right sort of time for a backup to get involved.

[callforjudgement]

Anyone know why the SE queue is quite so long, anyway? People wanting to play with newbies in particular? People who think 2of4 is a

really great

setup? People who think they aren't very good at playing still and want games with ICs in where they won't be shouted at?

[callforjudgement]

RTR's meta is all over the place; the rest of the site is rather more homogeneous.

This is almost certainly a direct consequence of putting 6 newbies in every game. (And not one that should be surprising, or discouraged.)

[callforjudgement]

Pretty much all ICs do have games outside. It takes far too long to get through five RTR games unless you do a lot of replacing in or you /in as SE three times in a row.

[callforjudgement]

The rules definitely used to be "mods cannot extend deadlines to beyond 3 weeks regardless of the circumstances". That seems to have warped into "deadlines must be 3 weeks exactly" at some point.

[callforjudgement]

SE shortages seem unlikely any time soon.

We do, however, get periodic IC shortages. (In fact, it'd seem likely to happen instantly upon Nacho going V/LA, which might happen at any time at random.)

[callforjudgement]

@Vi: The idea isn't expressed very clearly in the post, but it's that you cannot play in a newbie slot with 2 or more games of experience, and cannot play in an SE slot with 2 or fewer games of experience. So your third game has to be outside the newbie queue.

I'm not sure what I think of it myself.

[callforjudgement]

If you wanted to put a Roleblocker in with a JK, the only really good option is to consider a JK a "protect" ability for the purposes of NAR. (Therefore allowing it to be roleblocked) IMO it also makes sense with the way a roleblocker is meant to work.

[cfj-nar-rant]

NAR DOES NOT WORK LIKE THAT

[/cfj-nar-rant]
Please. If you're going to be discussing action resolution, get your terminology straight. Use NAR, or don't use NAR, but be straight on what it means. (Click the link for the CFJ NAR Rant explanation.)

(Especially in newbie games, where correctly explaining the rules is more important than usual.)

[callforjudgement]

I seem to remember they were all townsided except doc/townie. (Although my actual experience with cop/townie is that it's "balanced" in terms of win rate but very swingy, because it almost entirely depends on when the cop dies.)

[callforjudgement]

That's a big "if". Town are pretty much favoured if they get a scum lynch in the first two days even in a mountainous setup, IMO. It's just that they probably won't.

[callforjudgement]

A couple of us on another site playtested CES's matrix in real time mafia cause we were bored. Albeit, we tested with relatively experienced players, but still. (I generated BP/JK/Maf RB)

Most of them thought it was townsided, mostly owing to the nature of being able to better predict what other PR is in the game if you are a town PR. They found it enjoyable though.

Setups that are balanced in forum mafia are townsided in face-to-face Mafia, and setups that are balanced in face-to-face Mafia are scumsided in forum Mafia. So your results are encouraging; if it had seemed balanced in that test, it'd probably have been scumsided in practice.

[callforjudgement]

It's because it's much easier for scum to hide in forum mafia. Face-to-face is the most townsided because even if someone's not saying anything at all, you can still look at them to try to read their face or body language. Chatroom is less townsided, but more so than forum mafia, because in forum mafia you can randomly disappear for two days, turn up when you're prodded, and continue on like nothing else had happened; that hurts town no matter which alignment does it, and whether it's intentional or accidental. Also forum mafia easily degenerates into warring walls of text with half the players not contributing, and just generally being a pain to read; in realtime Mafia, if someone's not contributing, you can ask a question and expect a response within a few seconds. (That's another way realtime Mafia is more townsided: see DGB's sig.)

And yeah, a setup that's scumsided in face-to-face is almost certainly broken.

[callforjudgement]

The problem with static setups is they bring fakeclaiming down to a pure theory matter. So you end up with situations where players have to memorize a list of "what to claim when" and just follow it, which is not what we want to be teaching newbies. (Exception: mountainous setups, but those are a bad idea for other reasons.)

[callforjudgement]

When I'm ICing as town, I give scum advice right there in the thread. That way, the townies learn how to play scum too. (I've never been scum as an IC, but I'd do it in-thread then too.)

[callforjudgement]

For what it's worth, I've ICed one game, and afterwards, I didn't really feel like I could cope with being an IC.

Part of it's just the expectations problem from the newbies; they expect you to be a lot better than you are, which sort of drags you down into a burden of proficiency minefield you might not be used to.

Also people will try to interpret trying to help them as a sign of weakness. (Or worse, a scumtell.) This makes it quite hard to play like you normally would, which also makes it basically impossible to look town or scumhunt. So it's not really that fun to play in an IC role; you're just being focused on and dissected too much (when it should be the other way round!). I did enjoy the postgame analysis, though.

[callforjudgement]

I think I might have just found a partial breaking strategy for Matrix6?

At the start of day 1, any town Tracker claims. If such a claim is made (meaning this is setup C or 3), the other town power role claims non-vanilla, but doesn't claim the details of their power role. Players counterclaim if they hear someone claim something impossible (note that a scum Tracker claim is

always

counterclaimable, by the JK in setup A or 1, Cop in setup B or 2, and the real Tracker in setup C or 3). This inevitably either leads to a day 1 1v1 (which is definitely beneficial to town), or else two confirmed townies, and town have power roles on top of that. (Additionally, scum won't know who to shoot; if they aim for the Tracker they'll run the risk of hitting a Doctor protection, and if they aim for the other PR they might be Bulletproof. Either way, the scum kill is probably being directed away from strong scumhunters, giving town a large advantage; it's almost like they get to play in a 2:5 with no kill N1 or N2, except they have the benefit either of a Tracker shot or of a hard-to-kill unkillable confirmed townie, and that's even if scum guess

right

about who to kill.)

If everyone's verified to not be a Tracker, then it's probably best to play as normal (although I can see an argument for using a similar procedure to the above in setups A and 2 by getting the Doctor/BP to claim, it probably isn't worth it due to the roleblocker; that said, the smaller lynchpool would improve the chance of lynching the roleblocker and almost guaranteeing a town win in setup 2). Nonetheless, town will get a small advantage from the additional information they have about the setup (there's a 50% chance scum knew anyway, but even if they didn't, the information probably helps town more).

[callforjudgement]

There's a 50% chance there'll be a Doctor protection on the Tracker, though. When that happens, it guarantees that the Tracker will get off two investigations and survive until mylo. So shooting the Tracker is a pretty risky move from scum's point of view. The reward is mainly in the dayplay, rather than the nightplay (you're confirming two players, thus shrinking the lynch pool, and you're also making it hard for scum to shut down strong townies with their kill); the nightplay stuff is just a bonus.

[callforjudgement]

1. If they were, they'd be shot overnight anyway, so you aren't losing anything.
2. What you gain is a lynch from a pool of 7 on D1, followed by a lynch from a pool of 6 on D2, and a lynch from a pool of 5 on D3 (as opposed to the normal 9, 7, 5); additionally, if you hit scum on day 1, the pools are 7, 5, 3, 2 (compared to 9, 7, 5, 3) and if you hit scum on day 2, the pools are 7, 6, 4, 3 (again compared to 9, 7, 5, 3). This is because the Tracker is not only confirmed, but can also confirm townies without fear of dying. A D1 scum lynch when the Tracker is claimed and protected is almost unbeatable for scum. That's never going to happen if you don't claim. It's sometimes going to happen when you do.
3. I agree it's a bad way to introduce new players, but that's a fault of the setup, not the strategy itself. The claims affect the dayplay because, obviously, you don't lynch a confirmed townie (also, you know that their stated opinions are genuinely held). Innocent Child is a considerably stronger role than VT in terms of dayplay. Also, the statistics are quite clear that in Nightless games (where scum can't control who dies overnight), town winrates are

much

higher than predicted by EV theory; in games with a normal day/night cycle, town's success is comparable to chance or even worse. That's pretty strong evidence that the ability to control a kill is very valuable for scum.
4. I'm not quite sure what you mean by that.

[callforjudgement]

Has the following hypothesis ever been considered:

Surviving is an easier concept to grasp for a newbie, as opposed to scumhunting. Beyond that newMafia have a PT with which they can openly ask how to play and openly communicate with another player whose alignment is known to be the same as theirs. There's no paranoia for newScum.
In theory, the same should be true of newTown with regards to the IC, but that never happened in all the three newbie games I've been in.

Therefore, I put it to you that any setup that has a 50/50 winrate in the newbie queue will in-fact be sided towards town, and beyond that 'game balance' perhaps isn't so important.

Well, there was the experiment with giving scum daytalk (i.e. the ability to use the scum PT regardless of whether the game was in a day phase or a night phase) in Newbie games. For a while, they won almost every game, although it's widely suspected here that that was at least partly a statistical anomaly.

I think that there's another issue at play, though, rather than just communication issues. In setups where the scum have a relatively free choice of who to kill (such as Matrix6, given that the most commonly seen strategy is to try to keep the town power roles hidden), the scum nightkill is a very powerful weapon when the players differ in strength, as it makes it possible to eliminate the strongest Town players and leave the weakest players alive. (The clearest example of this is to take the following two setups which have similar results from random lynching: 11 Vanilla Townies versus 2 Mafia Goons, with a standard kill; and 7 Vanilla Townies against 3 Mafia Goons, with no nightkill. Both of these are predicted to have a town win rate of around 40% if players lynch randomly. Current belief is that the former setup is too scumsided to run – it's

never

been won by town on mafiascum.net – and that the latter setup is townsided.)

Now, in most games, we can assume that there's a spread in skill among the players. In the Newbie Queue, though, that's especially true; at least one player (the IC) is chosen specifically for their experience and has requirements to make sure that they have a sufficiently high skill level to teach the game well; meanwhile, new players vary wildly in how skilled they are, as do SEs (some SEs are highly experienced players just looking for a simple game, others are players who feel that they're not very good and aren't ready to play in the other queues). So games with a normal day/night cycle are likely to show more scumsided results in the Newbie Queue than elsewhere.

That said, it's unclear what it means for a setup to be "balanced". In general, I think you have to balance setups against the sorts of players that are likely to play them. The game will be very demoralizing for town if they need to randomly draw a "perfect team" in order to win. (The game's also unfortunately rather demoralizing for scum if a good player replaces into a bad player's town slot, as that can completely change the dynamics of the game and disrupt the strategy they were using; part of the reason the nightkill exists is to give scum a way to counter that.) So if scum have the strategizing advantage, we need to give town a little extra help in order to ensure that both teams have a fair chance. (And of course, a major part of the game is that scum are

meant to have

an information advantage; that's one of the main defining factors of how scumteams work in Mafia. So if the advantage turns out to be larger than expected, we can give town an advantage that fits their strengths to compensate, such as extra members or power roles.)