Town won 46% of Matrix 6 games

[TheButtonmen]

Also the current spree of posts wanting to pile punishments on those who replace out are being silly and/or dumb.

We're already chronically short on IC's, adding further disincentives for people to signup to IC just makes things even worse without actually helping to solve anything. The carrot rather then the stick would help a lot more in this situation.

[Mr. Flay]

We're really off-topic for this thread, so I have made a new one in Site Ideas: http://forum.mafiascum.net/viewtopic.php?f=7&t=37312

[Toomai]

Okay so I've done stats on the first 30 Matrix6 Newbie Games now (1370 - 1399), which is the first 35% of the 84 finished games mentioned in the first post (game starts between May 6th and July 3rd), and here are the major findings so far:

Spoiler: GRAPHS

I did not count replacements if they were pregame or noted by the mod as having "not picked up their role PM". Force-replacements were considered on a case-by-case basis - generally, if the player did it to themselves, it counts, but if it's some other person's fault, it doesn't. Modkills are ignored.

You can download the Excel file here. It will auto-update every time I change it, so redownload if you want to see the latest.

Things that jump out at me:

• Random is random, but that's a lot of setup 1s.
• While I haven't gotten to the 46% town winrate of the first post yet, and so this should be taken with a decent amount of salt, town is currently curbstomping scum in setups 1 and 2 whereas it tends to lose in the other four setups.
• The difference in replacement rate between town and scum is irrelevant for SEs, but noticable for newbies and ICs. It could just be bias right now for ICs (only 30 samples), but for newbies (150 samples) it's telling.

[BROseidon]

What if I have a death in the family, is that a good reason to replace out?
What if out of nowhere my boss challenges my work performance, " " ?
What if I break up with my girlfriend, " " ?
What if I'm severely frustrated by the game, " " ?
What if I'm just frustrated by the game, " " ?
What if I'm not having fun anymore and want to fuck off from this game, " " ?

Etc etc, obviously the " " in each example is just "is that a good reason to replace out"

Yes
Yes
Maybe
No
No
No

[BROseidon]

tl;dr See sig.

There's a certain irony that the person who said that replaced out of his first game on site.

[Mr. Flay]

I've done it too, a handful of times. I just don't think there should be zero consequences for it.

[Mr. Flay]

Random is random, but that's a lot of setup 1s.

I wonder if Setup 1 tends to finish faster somehow? AFAIK the List Mod is still randomizing the setups before giving them to the gamemod, so the previous bias should be gone.

[Zachrulez]

Random is random, but that's a lot of setup 1s.

I wonder if Setup 1 tends to finish faster somehow? AFAIK the List Mod is still randomizing the setups before giving them to the gamemod, so the previous bias should be gone.

Either that or sample size. You need more games with 6 possible setups vs 4 to balance out random chance.

[Plessiez]

Random is random, but that's a lot of setup 1s.

Surely not statistically significant though? I mean, the chances of

at least one

of the six possible set ups being chosen 10 or more times out of 30 can't be that low.

[Toomai]

Random is random, but that's a lot of setup 1s.

I wonder if Setup 1 tends to finish faster somehow? AFAIK the List Mod is still randomizing the setups before giving them to the gamemod, so the previous bias should be gone.

Actually it's the opposite. Setup 1 runs the slowest.


In any case, setup 3 is catching up in frequency as I continue to log. I agree it's probably not statistically significant.

[emeraldemon]

Thanks a lot for these stats Toomai! Are you recording each setup by hand? Are you keeping the data somewhere?

I would be willing to help with the logging if you can tell me what exactly you are recording. I still think with a little bit of effort and some clever regexes a bot could probably figure out a lot of these stats faster than a human could, or at least help with the data collection. Then maybe we could keep the info on the wiki somewhere?

[Sakura Hana]

Speaking of which I remember that i've replaced into newbie slots in newbie games 4 times and 3/4 of those times it was into scum slots.

[Egg]

What if I have a death in the family, is that a good reason to replace out?
What if out of nowhere my boss challenges my work performance, " " ?
What if I break up with my girlfriend, " " ?
What if I'm severely frustrated by the game, " " ?
What if I'm just frustrated by the game, " " ?
What if I'm not having fun anymore and want to fuck off from this game, " " ?

Etc etc, obviously the " " in each example is just "is that a good reason to replace out"

1. I'd take V/LA personally, but wouldn't hold it against someone who replaced out.
2. If playing mafia is the reason for this (I.E posting at work, games taking up all of your focus), yes. But replace out of ALL games and don't sign up for more.
3. No. Take a V/LA if you need to.
4. No.
5. No.
6. No.

[Toomai]

Thanks a lot for these stats Toomai! Are you recording each setup by hand? Are you keeping the data somewhere?

I would be willing to help with the logging if you can tell me what exactly you are recording. I still think with a little bit of effort and some clever regexes a bot could probably figure out a lot of these stats faster than a human could, or at least help with the data collection. Then maybe we could keep the info on the wiki somewhere?

I'm keeping the data in this Excel file (the link updates itself every time I change it). The first sheet shows what I'm recording.

I fit it a bit daunting to try and make a bot to record this, since there's judgement calls over whether a replacement should be counted or not, and cases where backup mods come in so you can't just look at only the first post's ISO. It only takes a few minutes per game anyway.

[GreyICE]

Random is random, but that's a lot of setup 1s.

I wonder if Setup 1 tends to finish faster somehow? AFAIK the List Mod is still randomizing the setups before giving them to the gamemod, so the previous bias should be gone.

No, it's that humans are really shitty at recognizing randomness.

Like, splitting 30 games between 6 groups (average 5), humans would say "random" would look like 7,4,6,5,3,5

10,4,6,4,3,3 and we go "why is 10 so big?"

Actually group 1 is laughably non-random and group 2 looks a lot more like what random chance will tend to produce.

[Toomai]

Okay I now have stats for the first 48 Newbie Matrix6 games (a bit more than half) (games 1370 - 1417).

Spoiler: MOAR GRAFZ

So the IC replacement rate isn't as bad as we thought it was. But the Newbie stats tell the story -

a newbie is 1.5 times more likely to replace out as scum

. The lack of parity for setups 1 and 2 is also evident.

I'll try to post once more once I get through all consecutive completed games.

[emeraldemon]

Hard to see what 1 and 2 have in common though. If you split by PR you get:
Jailkeeper = 12/18 = 67%
Cop = 8/12 = 67%
1-Shot BP = 5/18 = 28%
Tracker = 6/18 = 33%
Doctor = 8/12 = 67%

Small sample size still, but maybe 1-shot BP is weaker than other PRs?

[T-Bone]

Holy hell set-ups 1, 2, and 3.

[Alduskkel]

I am not convinced the results are statistically significant.

@emerald: Of course 1-shot BP is weaker than the other PRs. It can only help itself and only once.

[GreyICE]

I am not convinced the results are statistically significant.

I'm fairly confident they are.



That gives me p=0.01374, which I'd call significant any day of the week. So unless I fucked my math all to hell (possible) that's statistically significant on most levels we give a shit about.

[Alduskkel]

Sorry, I meant the balance statistics. I wasn't looking at the replacement statistics.

[kuribo]

The only games I've replaced out of involved cases where I ended up engaged in such a heated argument that I could no longer play to my win condition


(Pikmin mafia for example)


When it gets to a point that another player says I should be force replaced or the old "boot him or we quit" I'll generally replace out just to remove myself from the situation and avoid getting banned



This has only occurred in I think 3 games total, all three during my infamous issues with my cheating whore of an ex wife

[Jake from State Farm]

anyone track outcomes based on Day 1 lynch of town/scum/ and NL?

[Toomai]

anyone track outcomes based on Day 1 lynch of town/scum/ and NL?

[Jake from State Farm]

Wow thanks.