~~NewD3 Statistical Archive~~

[Somnus]

In another thread earlier this year, I said I'd give an update on NewD3 stats once it reaches 150 completed games. We have arrived. I think if there's interest, once in a blue moon, I'll bump this thread with random statistical developments.

Overall Results

Town Wins: 80 (53.33%)
Mafia Wins: 70 (46.67%)


A1: Town leads 10-8 (55.56%)
A2: Town tied 9-9 (50.00%)
A3: Town leads 9-6 (60.00%)
B1: Town leads 9-6 (60.00%)
B2: Town leads 6-5 (54.54%)
B3: Town leads 9-6 (60.00%)
C1: Town tied 9-9 (50.00%)
C2: Town leads 10-8 (55.56%)
C3: Town trails 9-13 (40.91%)

Town leads in 6 blocks, are tied in 2, and trails in 1. Mafia is only ahead in C3, which, coincidentally, has been rolled the most times out of any setup (22). B2 has been rolled the least (11).

Results by Column

Column A: Town leads 28-23 (54.90% and rolled 51 times)
Column B: Town leads 24-17 (58.54% and rolled 41 times)
Column C: Town trails 28-30 (48.28% and rolled 58 times)

Win-Streaks

Mafia has the longest overall win-streak in NewD3 at 6 (Newbie 2011-2016). As of now, that's the only time Mafia has ever won more than 3 in a row. Town has had 6 different win-streaks of more than 3, but have never won 6 straight. Town's longest win-streak is 5 (Newbie 1971-1975 and Newbie 1987-1991). The last time either alignment won more than 3 in a row was Newbie 2027-2030 (4 straight wins by Town).

2022 Results

Newbie 2085 and 2086 were ongoing at the start of 2022, so they will be counted as 2021 games (both were won by Town). There's been an interesting trend this year where Town has lost ground in the columns that they're overall ahead in. Meanwhile, Mafia is losing ground in Column C this year.

Starting with Newbie 2087 (the first Newbie game fired this year), Town trails Mafia 6-7.

Column A: Town has gone 2-4
Column B: Town has gone 1-2
Column C: Town has gone 3-1

The last 3 Newbie games were all in Column A. The 3 before that were all in Column C. Currently, Mafia is on a 3 game winning-streak (all in Column A) which is tied for their 2nd longest winning streak in NewD3. Until somewhat recently, Column A was 1 town victory away from pulling ahead of Column B for the most "town-sided" column. However, 3 straight mafia victories, all in Column A, have derailed that from happening. None of the last 7 games have been in Column B.

B1 and C3 have not been rolled this year.

Everything in this post is accurate (hopefully!) as of August 29th, 2022. It includes data from Newbie 1950 through Newbie 2099. Further statistical updates will not be edited in this post, and instead will be represented in future posts in this thread.

[Aisa]

I think if there's interest, once in a blue moon, I'll bump this thread with random statistical developments.

Yes please! I think it’d just be nice to continue to monitor winrates for individual setups.

[Irrelephant11]

A1: Town leads 10-8 (55.56%)
A2: Town tied 9-9 (50.00%)
A3: Town leads 9-6 (60.00%)
B1: Town leads 9-6 (60.00%)
B2: Town leads 6-5 (54.54%)
B3: Town leads 9-6 (60.00%)
C1: Town tied 9-9 (50.00%)
C2: Town leads 10-8 (55.56%)
C3: Town trails 9-13 (40.91%)

Every column being between the 40-60% range seems like an awesome outcome, and I have no problem with the fact that town wins a bit more than scum (I think it's been said many times that that's ideal for newbie retention)

[PenguinPower]

Thanks for putting in the effort to create this, Somnus. I'm actually quite pleased to see how evenly it's balanced overall, and it's a significant improvement over 2d3. Thanks to RadiantCowbells for the design.

It is interesting to see the disparity in the mason-setup winrates (A3/C3).

[Irrelephant11]

It is interesting to see the disparity in the mason-setup winrates (A3/C3).

Yeah that's funny to me

Added together they're 18-19, almost exactly 50%

[Alianna]

It is interesting. A3 and C3 are functionally the same as the roleblocker does nothing against masons, so I really don't see why there's a 19% difference in winrate.

[MegAzumarill]

It is interesting. A3 and C3 are functionally the same as the roleblocker does nothing against masons, so I really don't see why there's a 19% difference in winrate.

My guess is something psychologically on the side of mafia.

In setup A3 they have to be wary of : Cop, Doctor, Jailkeeper, and Tracker.
In Setup C3 they only have ot worry about jailkeeper and cop.

My guess would be that mafia feel more comfortable killing strong town players and strong townreads in C than in A, as opposed to roleblocking them, since it's twice as likely a protective exists in A. Killing strong townreads has good odds of hitting masons, while rbing them won't help.

In essence, in A mafia have an additional, safer tool to use than just their kill. This tool does nothing to masons, but that doesn't stop mafia from changing their play around said tool.

Additionally, a goon flip in C doesn't tell the town which column they are in, and lets mafia have more options of who to bus without accidentally making a solved game in rows 1 or 2.

[Somnus]

The A3/C3 thing was understandably a topic of interest when I first went through all of those Newbie games back in February. The conclusion that was drawn had to do with how Mafia have to play around protecting the Roleblocker. I had also noted at that point that we're still dealing with relatively small sample sizes. Getting a sample size of 150 for an Open Setup would be absolutely amazing, and while it's still a good sample to work with for a Semi-Open like this, we're essentially dividing that number by 9 since it's really 9 separate setups (150/9=16.67 times on average each block of the grid has been run).

So here's the thing: NewD3 was being absolutely dominated by Town through the first like...60 or so games. And that included A-3, where Town won 5 of the first 6 games in that block. At one point, Town was winning 2 out of every 3 games overall or so. Maybe higher. And then somewhere around Newbie 2010, Mafia started staging a comeback in NewD3 overall. It's always going to be a little town-sided. Once some of these other blocks in the grid, particularly in Column B where Mafia is at a somewhat noticeable disadvantage (in my opinion) start catching up with Column C as far as how many times they've been run, I expect NewD3's Town win% to go up a couple more points.

But having spent at least a little bit of time exploring all of the prior Newbie Game setups, this seems like far and away the best option. Stuff like 2 of 4 and Matrix6 had serious problems that could be exploited. NewD3 doesn't. Column B kinda sucks for Mafia a little bit, but I still wouldn't change it today if I could. Mafia Rolecop is a badass role, but in a game that typically ends on Day 4 (exactly 80% of games end on either Day 3 or Day 4), it's not a very powerful role here unlike Mafia Roleblocker. On top of all of that, the 10 roles that you can possibly get in NewD3 are just fun in general. NewD3 doesn't provide a ton of opportunities for Mafia to safely fake-claim, but I don't think that's really a terrible problem in Newbie Games. There's still been 3 or 4 successful fake claims in the Newbie Queue this year that I can recall off the top of my head (2088 and 2092 immediately come to mind).

As more time passes, I expect the Mason setups are going to continue to take a bit of a hit. Since I joined the site in January, there's only been 2 Mason games run in the Newbie Queue. Both were in A3 and both were won by Mafia (one of them was my first game!). So like...I think there's some legitimacy in the argument that the two blocks play out differently due to the Roleblocker, even if their night action ends up being effectively useless, but also note that Mafia has been catching up in that block after Town got off to a dominating lead.

I'm glad there's interest in this. The Newbie Queue is the only one where you have a fixed setup like this, so it's the only queue that lends itself well to tracking this kind of data.

[Somnus]

Another fun with numbers example: what Town PRs win the most often in NewD3? The answer might surprise you!

Friendly Neighbor: 15-11 (57.69%)
Doctor: 19-14 (57.58%)
Tracker: 27-21 (56.25%)
Jailkeeper: 25-22 (53.19%)
Cop: 19-17 (52.78%)
Mason: 18-19 (48.65%)

Worth noting, however, that the Friendly Neighbor only appears in Column B (B1 and B2) where, as previously mentioned, Town is at their strongest.

[Aisa]

I guess something like Meg's theory is completely plausible, but I am here to stress that

we're still dealing with

relatively

small sample sizes

[furtiveglance]

In another thread earlier this year, I said I'd give an update on NewD3 stats once it reaches 150 completed games. We have arrived. I think if there's interest, once in a blue moon, I'll bump this thread with random statistical developments.

Overall Results

Town Wins: 80 (53.33%)
Mafia Wins: 70 (46.67%)

A1: Town leads 10-8 (55.56%)
A2: Town tied 9-9 (50.00%)
A3: Town leads 9-6 (60.00%)
B1: Town leads 9-6 (60.00%)
B2: Town leads 6-5 (54.54%)
B3: Town leads 9-6 (60.00%)
C1: Town tied 9-9 (50.00%)
C2: Town leads 10-8 (55.56%)
C3: Town trails 9-13 (40.91%)

Town leads in 6 blocks, are tied in 2, and trails in 1. Mafia is only ahead in C3, which, coincidentally, has been rolled the most times out of any setup (22). B2 has been rolled the least (11).

Results by Column

Column A: Town leads 28-23 (54.90% and rolled 51 times)
Column B: Town leads 24-17 (58.54% and rolled 41 times)
Column C: Town trails 28-30 (48.28% and rolled 58 times)

I thought I'd update this now that we've done 160 games.

Overall Results

Town wins: 84 (52.5%)
Mafia wins: 76 (47.5%)

A1: Town leads 11-8 (57.89%)
A2: Town tied 9-9 (50.00%)
A3: Town leads 9-8 (52.94%)
B1: Town leads 10-7 (58.82%)
B2: Town leads 6-5 (54.54%)
B3: Town leads 9-6 (60.00%)
C1: Town tied 9-9 (50.00%)
C2: Town tied 10-10 (50.00%)
C3: Town trails 11-14 (44.00%)

Town leads in 5 blocks, are tied in 3, and trails in 1. Mafia is only ahead in C3, which, coincidentally, has been rolled the most times out of any setup (25). B2 has been rolled the least (11).

Results by Column

Column A: Town leads 29-25 (53.70% and rolled 54 times)
Column B: Town leads 25-18 (58.14% and rolled 43 times)
Column C: Town trails 30-33 (47.62% and rolled 63 times)

[T-Bone]

Thanks, Furtive!

[Thestatusquo]

The real question becomes when do we start using these numbers to wager on upcoming newbie games?

[MegAzumarill]

Wait we weren't doing that already?

[Thestatusquo]

An outrage I think personally.

[Cook]

hey folks! i'm in the process of updating this, seeing as newd3 as a dedicated queue is coming to a close. i'll be expanding on this work to include more statistics and database analysis, which i'll then publish afterwards

[Cook]

i may also analyze the other 1949 newbie games that have come before, and see just how different things are. that's a down-the-line thing though

the newbie queue is a neat capsule of setup balance and design that i think is unique to MS as a site and i think we deserve to have some analysis done on it to see what we could do in the future

[Alianna]

hey folks! i'm in the process of updating this, seeing as newd3 as a dedicated queue is coming to a close. i'll be expanding on this work to include more statistics and database analysis, which i'll then publish afterwards

You beat me to it! Was going to say, it's definitely a good time for an update. Looking forward to the results.

[Ausuka]

Hi - I did what I believe to be an accurate update here, up to the end of the Newbie queue;

Town wins: 95 (52.8%)
Mafia wins: 85 (47.2%)

A1 (Cop + Doc vs RB): Town leads 12-9
A2 (JK + Trk vs RB): Town leads 12-9
A3 (Mason x2 vs RB): Town leads 11-9
B1 (Trk + FN vs RC): Town leads 12-9
B2 (JK + FN vs RC): Town leads 8-6
B3 (Trk + Doc vs RC): Town leads 10-6
C1 (Cop): Mafia leads 9-11
C2 (JK): Mafia leads 10-11
C3 (Mason x2): Mafia leads 11-15

Column A: Town leads 35-27 (town winrate 56.5%)
Column B: Town leads 30-21 (town winrate 58.8%)
Column C: Mafia leads 30-37 (town winrate 44.8%)

[Ausuka]

This is a fairly simplistic update and I hope it doesn't overshadow what cook is doing - I wanted to do this anyway and thought it would be better to share it than not