Updated NewD3 Win%

[Somnus]

This took hours over the last few nights. This includes every completed game in which NewD3 has been used. Currently, there have been 138 (Newbie 1950-2087). I plan on updating this after Newbie 2099 when the sample size is 150 games. Overall, pretty pleased with the results. Enjoy.

Overal NewD3 Games: 138
Town: 74 wins (53.62%)
Mafia: 64 wins (46.37%)

A1 Games: 16
Town: 9 wins (56.25%)
Mafia: 7 wins (43.75%)

A2 Games: 16
Town: 8 wins (50%)
Mafia: 8 wins (50%)

A3 Games: 14
Town: 9 wins (64.28%)
Mafia: 5 wins (35.72%)

B1 Games: 15
Town: 9 wins (60%)
Mafia: 6 wins (40%)

B2 Games: 10
Town: 6 wins (60%)
Mafia: 4 wins (40%)

B3 Games: 13
Town: 8 wins (61.54%)
Mafia: 5 wins (38.46%)

C1 Games: 15
Town: 7 wins (46.67%)
Mafia: 8 wins (53.33%)

C2 Games: 17
Town: 9 wins (52.94%)
Mafia: 8 wins (47.06%)

C3 Games: 22
Town: 9 wins (40.91%)
Mafia: 13 wins (59.09%)

Grid Setups from Town Lean to Mafia Lean (by current win%)

A3 Town wins 64.28%
B3 Town wins 61.54%
B1 Town wins 60.00%
B2 Town wins 60.00%
A1 Town wins 56.25%
C2 Town wins 52.94%

A2 Town wins 50.00%

C1 Town wins 46.67%
C3 Town wins 40.91%

[Morning Tweet]

very nice!

[Morning Tweet]

A3 Town wins 64.28% Town wins 64.28%

C3 Town wins 40.91% Town wins 40.91%

The best setup for town is Mafia Roleblocker vs. Mason/Mason, whereas the best setup for mafia is Mafia Goon vs. Mason/Mason.

mechanically those two setups have no difference and the deviation is just random, no? pretty satisfying to see that no setup appears to have dominated.

Which is funny, I thought for sure the ones in which town autowins mechanically via a D1 scumlim would see a higher winrate. my guess is that town limming scum D1 is so good for newbie games to the point where it doesn't matter if mechanically scum loses after one -- they would have lost anyway

[Somnus]

That was my biggest takeaway as well. A3 and C3 are effectively the same setup. My only game so far, I was a Mafia Roleblocker and we NK'ed a mason night 1, which informed me that I was basically a glorified goon. So yes, definitely just random deviation due to a small sample size. A sample of 138 games isn't much, especially when split 9 ways.

So there's probably not any sweeping statements I can make on the balance of NewD3 given, as I said, a fairly small sample size. However, my humble opinion is that it's a great setup for newbie games (coming from someone who is pretty "meh" about semi-opens typically).

[RadiantCowbells]

So here's a fascinating question

Is the reduced winrate with Roleblocker on the team potentially caused by actual differences in scums approach to the game

[Somnus]

So here's a fascinating question

Is the reduced winrate with Roleblocker on the team potentially caused by actual differences in scums approach to the game

I was hoping you'd chime in!

It's a good question and I wish I had an answer. To date, I think there's a grand total of 3 newbie games that I've read start to finish (and one of them, I was in), so I can't say that I have or plan to scour through all of them in their entirety. My initial hunch would be to chalk it up to random deviation of what is still a rather small sample size.

So you're suggesting that perhaps mafia is playing worse when they have a roleblocker and have the threat of potentially facing cop+doctor or Jailkeeper+Tracker as opposed to knowing that masons aside, they're only up against a cop OR a jailkeeper?

[Ythan]

Was wondering that as well ego

[Somnus]

I think NewD3 is overall fairly well balanced, if not slightly town-leaning. And while ideally, you want the setup to be virtually 50-50 chance, considering they're newbie games, I don't think setups that are slightly town-leaning are bad for newbie retention. People are more likely to stick around if they are a part of a win on their first game, and the majority of brand-new players are going to be town (I wasn't town, but I digress).

Maybe that's a bad argument for me to make, but I think there's something to be said for retaining players if the setup is slightly town-leaning. The way I currently see it if you're town:

Column B: Easy mode
Column A: Normal mode
Column C: Hard mode

[RadiantCowbells]

My pursued winrate as a mod/designer is actually generally 55 town 45 scum but as a reviewer i'll pass anything from 60-40 town-scum to 45-55 town-scum

[Somnus]

Sounds reasonable.

[StrangeMatter]

One gripe I do have is that fakeclaiming PR just ends up being less than optimal as either scum or town. Often brings down games more often than not.

[RadiantCowbells]

That's intended functionality :p

[RadiantCowbells]

In Matrix6 scum who fakeclaimed diagonal PR won an awful lot of the time and i didn't like the amount of power that knowing correct fakeclaims gave scum and how that interacted with newbie newbie vs newbie experienced scum team

[RadiantCowbells]

You can be really bright and have great social deduction skills but if you're new you simply won't be able to figure out how why and when to fakeclaim without some sort of assistance so I felt like making a setup that doesn't really reward fakeclaims makes it more newb friendly

[MegAzumarill]

So here's a fascinating question

Is the reduced winrate with Roleblocker on the team potentially caused by actual differences in scums approach to the game

It does possibly in some scenarios.
Newbie 2085 was a game I played in where scum counterclaimed cop when I claimed it as mason.
This would almost certainly not happen if the game was goon/Goon.

As for general play I'm not one to say

[Not Known 15]

So here's a fascinating question

Is the reduced winrate with Roleblocker on the team potentially caused by actual differences in scums approach to the game

Yes. But that's actually part of the setup.
For two of the three setups in the roleblocker column losing the roleblocker early can be extremely bad. This means that all of the three setups have the + town EV part of scum protecting the roleblocker.
The column also has a doctor, which means that scum has to decide between the assumption that the claimed cop is a mason or the one that it is not. A wrong guess is again +ev for town, which drives up town winrate for both setups.
The two subsetups are not the same at all, and the winrates are probably correct.

[RadiantCowbells]

So here's a fascinating question

Is the reduced winrate with Roleblocker on the team potentially caused by actual differences in scums approach to the game

Yes. But that's actually part of the setup.
For two of the three setups in the roleblocker column losing the roleblocker early can be extremely bad. This means that all of the three setups have the + town EV part of scum protecting the roleblocker.
The column also has a doctor, which means that scum has to decide between the assumption that the claimed cop is a mason or the one that it is not. A wrong guess is again +ev for town, which drives up town winrate for both setups.
The two subsetups are not the same at all, and the winrates are probably correct.

good answer

this is 2/3 of what I believe the answer to be

[Not Known 15]

So here's a fascinating question

Is the reduced winrate with Roleblocker on the team potentially caused by actual differences in scums approach to the game

Yes. But that's actually part of the setup.
For two of the three setups in the roleblocker column losing the roleblocker early can be extremely bad. This means that all of the three setups have the + town EV part of scum protecting the roleblocker.
The column also has a doctor, which means that scum has to decide between the assumption that the claimed cop is a mason or the one that it is not. A wrong guess is again +ev for town, which drives up town winrate for both setups.
The two subsetups are not the same at all, and the winrates are probably correct.

good answer

this is 2/3 of what I believe the answer to be

Well, some people won't like that the newbies are exposed to (in this case legitimate and only slightly gambit-y) town fakeclaim strategies in a newbie game...

[StrangeMatter]

Isn't it weird that I think while WRs probably have a place in Mafia, logically speaking with a game like Mafia they really aren't going to be accurate?

[Somnus]

Not at home, so I don’t have the numbers in front of me, but interestingly enough, town won 5 of the first 6 A3 games, if I recall correctly. Then mafia staged somewhat of a comeback in A3 games.

As a whole, town was dominating to the tune of something like winning 2 out of every 3 NewD3 games through the first 60 or so. Then again, mafia started winning more to bring the winrates into reasonable territory. Make of it what you will. I should probably just link the google spreadsheet.

[RadiantCowbells]

Well, some people won't like that the newbies are exposed to (in this case legitimate and only slightly gambit-y) town fakeclaim strategies in a newbie game...

The difference is that fakeclaiming exists in this setup as a protown move only when the town players know all the town roles which is a situation that doesn't otherwise come up

[RadiantCowbells]

If you've ever played card games like Magic Yugioh or Hearthstone i think it's a good context to think of winrates in a mafia context. Better players / team compositions (roles and such) consistently win more but that's not the same as saying they are guaranteed to win, just that having better players or roles or other dynamics on a team raises your chance of winning. I don't think there is a more accurate tracker of performance than winrate.

[Somnus]

Of course. Better teams are going to win more often. I still believe it’s a rather small sample size, but I’m pretty pleased with the results overall. It is interesting to me that town flew out of the starting gate with a really high win percentage and then mafia staged a significant comeback. I’m not offering any sweeping statements other than presenting the data and seeing what people think of the results. I’m a sucker for this kind of stuff.

[RadiantCowbells]

Obviously the system becomes far more meaningful if you institute some sort of ELO system so that stacked teams are rewarded less and upsets are rewarded more, like the game that Who had the misfortune to play a two hour three person marathon game with myself and NSG shouldn't in a hypothetical universe affect much for him. Unfortunately something like that would never happen on this site, both because there aren't many people playing competitively anymore and because, well, a lot of people who convince themselves that they're still Top Tier players while having little concrete basis for thinking so would resent actual evidence of such. The current system where concrete evidence isn't taken as seriously as people's feelings about how good people are is very comforting for people who want to feel a certain way but aren't willing to actually put in the work, the soul searching, the meta analysis of ones own games it takes to actually improve. The way MS currently works is essentially who is good? Who is to say. I have my opinions, you have your opinions, the vast majority of people who think they are really good don't have good performances by any measurable standards so they disbelieve those standards and those standards become irrelevant because the majority has spoken, so the person who is good is actually the person who i like the most.