Newbie Data

[mith]

You'd expect some random drift, but it's still a bit suspect that the no-power-roles setup is the one that keeps coming up low, given the C9 trend. At least we're doing a better job there.

I think someone worked out an actual optimal strategy for that, but I don't recall it offhand. Faster game, but more random; I think F11 provides enough cover for a lone Goon to pull it out, at least occasionally.

Eh. The main difference in F11 is that half the time the setup won't have power roles, so the Goon has a chance in the resulting Vanilla game. In the setup with both power roles, it's probably a little worse for the Goon going into Night (less chance of hitting the Cop, one extra mislynch for the town). It

feels

a little better, though, because in the Pie E7 worst case, town has a forced win (D1 RB lynch, Cop and Doc both survive the night), while in F11 the scum still has some faint hope (I think). The extra townies hurt the town in that specific scenario (but overall, the EV is better for the town entering N1, if they have both power roles).

All that said, I've never had a problem with that aspect of Pie E7. The value of the RB means that the scum should play to make sure the RB stays alive, which potentially gives more grouping-tells, which makes for a better game. If the town hits the RB D1, it's because the scum played poorly (or because the town played really well); I'd guess the fact that more RBs were lynched than Goons on D1 is a result of the small sample size.

SensFan: It's worth noting that the Goon can't fake claim in F11 either, *even if there are no power roles* (again in the case where the RB was lynched D1), because no Doc will claim and the town will realize they aren't in the both-power-roles setup. (Correct play in F11 is for the Cop to come out D2, if there is one, and then for the Doc to come out D3, if still alive. This ensures that the Cop survives to the endgame, and confirms the Doc even if the Cop hasn't investigated him, and catches lying scum if he fakes in a no-power-role setup.)

[mith]

Why not just get the newbie list mod to do the selection?

We did this for a little while... it's a bit of additional work for the List Mod, though, and we shouldn't

need

to do it. (And it's clear with F11 that we have hammered home the "do it randomly" point enough that very few, if any, mods are "cheating", whereas with C9 it was a pretty significant problem. I don't have any reason to think it's affecting games.)

(It's worth noting for historical reasons that the disparity in the C9 numbers was at least partly due to mods using different methods for choosing the roles, rather than mods "rerolling" when they hit the vanilla setup. I remember someone, due to confusion over what "C9" stood for, randomized the setup by choosing 5 pro-town roles out of the 7 "available" roles - 1 Cop, 1 Doc, 5 Townies - and the probability of getting the vanilla setup doing it that way is only 1/21. Others did a 1/6 - 1/3 - 1/3 - 1/6 split, rather than 1/4 - 1/4 - 1/4 - 1/4. But I think we did have a few own up to throwing out the vanilla setup.)

[mith]

(The obvious modification to make to F11 to improve the town win rate - and I'm sure it has been suggested before, either for F11 or C9 - is to throw out the vanilla setup, of course. Mafia know if they are in a 2 power role setup, which makes it Pie E7+2, and they can protect the RB accordingly; and they know if there's only 1 power role, but not which, preserving the "power role hunting" aspect - in order to fake claim more effectively, they need the real power role dead or outed or figured out.)

[mith]

Oh, wow, I had only looked at the

C9

Pie E7 lynch numbers. That's... surprising.

[mith]

Sorry, I meant Pie E7 lynch numbers (regarding the frequency of Roleblocker lynches vs. Goon lynches).

[mith]

I'm not opposed to 1, but we can't seem to win 2:10 Vanilla Games with experienced players. Upping the size makes them last WAY too long (and essentially turns them into Minis), and keeping them at 2:5 or 2:7 makes them uncomfortably hard for Town to win.

To summarize from the Pie E7++ thread:

a. The 2:10 results are a very small sample size (and actually we do have one town win in a 2:8, even with that small sample).
b. Going to a 2:9 Vanilla would make them last about one game day longer on average - which may or may not be a terrible thing - and of course, we could take measures to reduce the day length, if game length is a big concern. [Actually, I expect just going to a Vanilla setup would reduce day length a bit on its own - no waiting on claims.]

We're somehow winning an average number of those, as Elmo pointed out in post 17, but it's probably through blind luck.

Well... yeah, on a large scale, probably so. If the newbies are doing about what the EV would predict, then they're doing as well as they would by lynching randomly,

by definition

. A phrase like "through blind luck" has the connatation of doing something really unlikely, though, and that's not the case here. It's not unlikely that the "sample mean" is close to the EV - it's the default assumption.

(On a game-by-game basis, I expect that skill does play an important role in the results - it's just that skill appears to be averaging out, for the most part.)

F11 set, Day One lynch data: of the 13 games where the Cop is the first lynch, Mafia win the game 92.31% of the time.

We should be careful with numbers like this. Note that if the town lynches the Cop, they could always just decide they suck and start lynching randomly the rest of the game (EV of about 30%) - and we actually don't have enough data to reject the null hypothesis of "Newbie towns win 30% of games after lynching the Cop D1". I would guess the 1/13 stat on this is partly bad luck and partly due to the fact that a town lynching a Cop D1 is a pretty awful town to begin with.

[mith]

FWIW, I wasn't suggesting we eliminate the vanilla part of F11 - I was only suggesting that that would be an obvious way to improve the win rate, if we are concerned about it being too low.

[I'm somewhat amused (but not surprised) that the suggestion of going to a vanilla setup brought out various posts about the importance of exposing newbies to power roles, while the suggestion of eliminating the vanilla part of F11 brought support for the vanilla setup as the best way to teach scumhunting values.]