The Newbie Setup and C9

[mikeburnfire]

[mrow][col]Games Played[col]Games Won by Town[col]Win %

Both

[col]62[col]22[col]35.48%

Cop only

[col]59[col]19[col]32.2%

Doc

[col]37[col]13[col]35.14%

None

[col]24[col]7[col]29.17%So even with no power roles at all, the town wins nearly a third of the games (which is better than they've managed in large-scale Mountainous Games). Whether that's due to newbie scum or other factors, I can't say, but I don't think it's irredeemably broken.

Hooray for evidence!

[Cogito Ergo Sum]

I blame ICs. If I can win, then so can all of you!

(I just won another one as town.)

[Yosarian2]

Actually, I tend to think that compent players on all sides would actually lead to an even MORE unbalanced game with more scum wins; if the scenerio is unbalanced in favor of the scum, then eliminating random factors (like random newb mistakes) would make the game more consistantly won by the side with the inherent advantage.

I'm not saying that noob mistakes are random factors, I'm saying that, in the long run a group of experienced town will do better against a group of equally experienced scum than they would if they lynched randomly.

And you don't think that's true with newbie town and newbie scum? I would think newbie scum would tend to give off bigger and move obveous scumtells then experenced scum, and odds are there's at least one IC townie who can pick up on them.

[Yosarian2]

I blame ICs. If I can win, then so can all of you!

(I just won another one as town.)

You're not a stasticially significant sample. Play a hundred newbies, check your win percentage, and then we can talk.

Anyway, I think Mr. Flay's stats pretty much destroy the theory that "4 town, 1 cop, 2 scum" games are unbalanced in favor of the town.

[mith]

I should point out that the "breaking strategy" for the original Newbie game only yields about 65% for the town with random lynches/choices. The point of switching to C9 was not so much that the game was horribly biased in favor of the town (2-5 is far more biased in favor of the scum from a random-lynch point of view... about 77%), but that the best play for the town was boring and didn't teach newbies much.

One possible factor in the Doc-Only > Cop-Only numbers is that scum are in general more likely to claim doctor (do we have any stats on that in newbie games?). It's also worth noting that the "Games Played" is nowhere near the 25-25-25-25 distribution intended; any stats on just more recent games, after we made an effort to make the distribution more clear?

[Fiasco]

One possible factor in the Doc-Only > Cop-Only numbers is that scum are in general more likely to claim doctor

That, and good old random noise.

[Fiasco]

One possible factor in the Doc-Only > Cop-Only numbers is that scum are in general more likely to claim doctor

That, and good old random noise.

To expand on this: standard deviation for a binomial distribution is root(npq) which (if p is about .3) works out to something on the order of 10% (of total games) for no power roles and something on the order of 5-10% for both power roles. So while the win probabilities for different setups probably aren't

too

far apart, I don't think there's much evidence for any other weird effects to explain like docs being more effective than cops or town winning too much with no power roles.

It's also worth noting that the "Games Played" is nowhere near the 25-25-25-25 distribution intended

Based on the data, I'd guess at most 1/3 of newbie mods randomize, with most of those being MeMe. (reference)

100% cop 50% doc may be a good setup in that it never screws the town over completely, but leaves room for setup uncertainty and risky fake claims.

[mith]

Oh, I agree that random noise is the main factor. I would also suggest that the win probabilities for the different setups *would* be much farther apart if they were open, and that the semi-openness of the thing is bringing them closer together. The doctor thing is just a theory.

Whether we stick with C9 or use something else semi-open, I'd be in favor of the List Mod randomizing the setup in the future (assuming the stats from after the pj thread show the distribution is still not consistent with randomization).

[Cogito Ergo Sum]

I blame ICs. If I can win, then so can all of you!

(I just won another one as town.)

You're not a stasticially significant sample. Play a hundred newbies, check your win percentage, and then we can talk.

I admit I've only played about 10 newbies as Town, so my experiences could be purely coincidental in nature, but I find that by asking lots of questions, playing responsibly and putting in some effort, newbie games are very much winnable.

(I just won an experienced C9 as town.)

[Mr. Flay]

mith, do you remember when the announcement was made that C9s had to be randomized? Going back a year.... errr, no. Still not really random (35% Both Roles, 33% Cop-only, 18% Doc-only, 15% No Role). There's really not enough games since the October discussion to get a good gauge, but going back just about 20 games, it's somewhat better: 32/32/12/24. I'd somewhat agree that the List Mod should send a random setup to the mod, though I'm pleased to see that my own stats are pretty close to randomized (not giving out numbers since I'm running some now).

Also, if anybody can do a chi-square on the total stats that might tell us something:
[mrow][col]Cop[col]Doc

Yes

[col]125[col]101

No

[col]63[col]87

[mrow][col]Games Played[col]Games Won by Town[col]Win %

Both

[col]62[col]22[col]35.48%

Cop only

[col]59[col]19[col]32.2%

Doc

[col]37[col]13[col]35.14%

None

[col]24[col]7[col]29.17%So even with no power roles at all, the town wins nearly a third of the games (which is better than they've managed in large-scale Mountainous Games). Whether that's due to newbie scum or other factors, I can't say, but I don't think it's irredeemably broken.

Hooray for evidence!

Yeah, but I just realized I'd misfiled one game as C9 that wasn't, so the numbers above are just slightly high now. The same ratio applies, though...

[Fiasco]

I wonder if from a teaching point of view, making newbies read a "dumb mistakes not to make" guide isn't at least as effective as making them play a newbie game.

Yet another possible C9 alternative: if there's no cop, turn one scum into a townie. (The remaining scum then knows there's no cop...)

[MrBuddyLee]

It might be helpful to pinpoint the most important lessons for town newbies to learn. Roughly in order:

1. Take your time. You need to hear everyone talk about everyone else in order to properly judge their sincerity and alignment, so don't rush to lynch just anyone. Town also needs to pinpoint a full matrix of player-player interactions each day to use as reference material the next day.

2. Don't help scum find the power roles. If you say you're not a power role or ask someone else if they are, you're giving scum information to work with.

3. Say more rather than less. Other townies will gauge your alignment by your words, and the more you say the tougher it is to fake your alignment. If you ask specific questions, you'll get answers or evasions, both of which are helpful. Sitting back and saying nothing is easily interpreted as not being interested in finding scum i.e., you are scum.

4. Don't overreact to people's mistakes. Be sure to take the time to ask why they made the mistake. Newbie town can make mistakes just as easily as scum.

5. Watch out for scum piling on with votes. Be aware of the vote count and what could happen worst case scenario if the scum haven't placed their votes yet. Be sure you don't give them an opportunity to finish the day prematurely, or be clear that if they do that they'll be your top suspect tomorrow.

6. Make mountains out of molehills. Days can't get started if you only talk about things that look REALLY important. You need to nitpick at stupid little things sometimes to get the game moving.

7. Don't necessarily trust a claimed power role. That cop... might not really be a cop. Enough said.

8. Look for players protecting each other. If someone doesn't find a suspicious player suspicious, ask why. If the explanation doesn't make sense, you may have found scum.

[MrBuddyLee]

Important lessons for newbie scum:

1. Don't give away your partner. Treat them like you'd treat any other player or people will spot the inconsistency. Even if you get lynched, your partner can still win the game for you, so don't bring them down with you.

2. Look for the doctor and cop. Sometimes people make mistakes that hint at their role, and open the door for you to squish them.

3. At least pretend to be looking for scum. Pretend you're town and find posts that look suspicious and attack them.

4. Vote wisely. Don't just pile on to lynch when you see the opportunity or you might get nailed tomorrow first thing. Be sure your votes are justifiable and fit into your master plan for winning the game.

[Fiasco]

Thanks, MBL. I'm probably going to quote that in any future newbie games. I would add this, to be precise:

"Don't put a third vote on someone unless you're sure you wouldn't mind a lynch. A careless third vote gives the mafia an opportunity to finish off one of your allies, and then you'll probably look pretty suspicious yourself.

Second votes are less dangerous, but keep in mind that they could give some foolish townie the opportunity to place a third vote."

[Cogito Ergo Sum]

First votes are even less dangerous, but keep in mind that a pair of foolish townies could place the second and third vote, giving the scum a chance to hammer.

Not voting is however completely harmless.

[Norinel]

Anyway, I think Mr. Flay's stats pretty much destroy the theory that "4 town, 1 cop, 2 scum" games are unbalanced in favor of the town.

If your primary argument against C9 is that town doesn't have a chance, why would you switch to something where the empirical evidence for almost the same case (Except that the town doesn't know there's a cop and no doc) gives the town about the same if not chance at winning? And if the town's information of exactly knowing the setup helps them enough to give the town good odds, how do you know the odds aren't so good that they make it imbalanced for the town?

FOS: Yosarian2

[Yosarian2]

(shrug) Well, the origional newbie wasn't unbalanced in favor of the town until the breaking stratagy was developed. I can't imagine that a weaker setup with only a cop would be unbalanced in favor of the town.

As for the emperical evidence, eh. The only reason the odds for c9 aren't even lower then they are is because the mods are apparently setting it up so that the first two cases (cop+doc and cop) happen a lot more often then the second two (just doc and no power roles). The only one that seems unexpected is the surprisingly high sucess rate for the doc only games, and with only 50 of those games played in total, that's probably a stastical anomoly; remember, 2 extra games won by the town out of 50 would be enough to push it up by 4%, which is the entire difference between where the doc games are and where I'd expect them to be.

[Fiasco]

More tips that I'd want newbies to read (that sort of overlap with MBL's):

"Yes, you are allowed to claim your role. However, there's no good reason to claim unless you're close to being lynched and you think it may save your life. You don't want the scum to know who to kill. Even knowing you're a vanilla townie is useful information to them.

Yes, you are allowed to make false claims about your role, but if you're town, it's generally a very bad idea. Trust me on this. Don't lie.

To judge whether you're town or scum, the town will need to see as much content from you as possible. If all you're going to post on a game day is three one-liners, there's not enough information to actually play mafia. Other than not wanting or being able to spend the time/energy, there is no valid excuse to ever withhold your opinions on anything related to the game. If you don't have any opinions, think harder. (I wish experienced players took this to heart, too.)"