how to study voting patterns?

This forum is for discussion related to the game.
User avatar
p2
p2
Townie
User avatar
User avatar
p2
Townie
Townie
Posts: 95
Joined: June 11, 2015

how to study voting patterns?

Post Post #0 (ISO) » Wed Jan 27, 2016 9:30 am

Post by p2 »

Hi guys,

I know that some of you have been interested in analyzing data about how alliances/associations might be detectable from voting patterns and whatnot. I've got a base of data collected over the years ( :( ) that I could probably expand to include information about people's votes over the next couple of days, but I'm having trouble conceptualizing how I might analyze it, which in turn makes it hard to conceptualize how I should organize and represent it.

So I guess I wonder - if you had your hands on a spreadsheet outlining ~100 games worth of voting data and the context surrounding those votes (ex. post number, day, voter/voted alliance, who's alive, etc),
what would you do with it
?

And perhaps more broadly for anyone else who cares about voting patterns but doesn't care to handle any data themselves -
what do you expect the data to demonstrate
?

A sanity test of the results might try to confirm that informational power roles do better than other townies at landing their votes on scum.

Another obvious thing to do might be to examine bussing - how often mafia vote for the buddies, how often they're actually on the wagons of their buds once they're finally lynched, etc.

But - as I actually don't play mafia very much any more - I expect that practitioners of VCA might believe or expressly disbelieve in other patterns of voting behavior that this sort of data might illustrate. It'd be nice if any of those practitioners would review their approach to VCA here.

Thanks.
User avatar
itlepip
itlepip
Mafia Scum
User avatar
User avatar
itlepip
Mafia Scum
Mafia Scum
Posts: 3076
Joined: September 21, 2015

Post Post #1 (ISO) » Wed Jan 27, 2016 11:11 am

Post by itlepip »

One thing to be aware of when you do this is that the alignment of each player effects who they vote for. Like if everyone voted randomly scum would still be less likely to vote for scum than town.
User avatar
Antihero
Antihero
al;kdjfal;kj
User avatar
User avatar
Antihero
al;kdjfal;kj
al;kdjfal;kj
Posts: 15872
Joined: March 30, 2009

Post Post #2 (ISO) » Wed Jan 27, 2016 12:15 pm

Post by Antihero »

In post 0, p2 wrote:Another obvious thing to do might be to examine bussing - how often mafia vote for the buddies, how often they're actually on the wagons of their buds once they're finally lynched, etc.

part of this is tied to individual scum styles (some people tend to bus more than others) and part of this is tied to whatever the site meta is at the time. i would expect this rate to vary quite a bit.

vca itself isn't a very good scumhunting tool.
The distance between insanity and genius is measured only by success.
User avatar
RadiantCowbells
RadiantCowbells
He/him
Smooth Criminal
User avatar
User avatar
RadiantCowbells
He/him
Smooth Criminal
Smooth Criminal
Posts: 70855
Joined: February 24, 2013
Pronoun: He/him
Contact:

Post Post #3 (ISO) » Wed Jan 27, 2016 12:17 pm

Post by RadiantCowbells »

Go read NY184's votecounts on Day 1 and 2.

Color in both EPM and Lost as scum, then the three flipped town as green.

GL VCAing that.
2019 stats: Town WR 76.7%, overall WR 81.667%, 1 scum defeat involving a major mod error in lylo vs 8 scum wins.
User avatar
Psyche
Psyche
he/they
Survivor
User avatar
User avatar
Psyche
he/they
Survivor
Survivor
Posts: 10058
Joined: April 28, 2011
Pronoun: he/they

Post Post #4 (ISO) » Wed Jan 27, 2016 12:19 pm

Post by Psyche »

http://forum.mafiascum.net/viewtopic.ph ... s&start=50

forgot this thread
suggests that this sort of study won't be worth the time
might be better to focus on tone variation then!
but even that might prove useless
youtube playlist extracter | donbot | game scraper | vca | setupsim | strategist | llm
User avatar
BROseidon
BROseidon
Expert Marxman
User avatar
User avatar
BROseidon
Expert Marxman
Expert Marxman
Posts: 8242
Joined: April 18, 2013

Post Post #5 (ISO) » Wed Jan 27, 2016 12:23 pm

Post by BROseidon »

I definitely use VCA with some ~intuition~ to try to structure the game sometimes.
User avatar
Psyche
Psyche
he/they
Survivor
User avatar
User avatar
Psyche
he/they
Survivor
Survivor
Posts: 10058
Joined: April 28, 2011
Pronoun: he/they

Post Post #6 (ISO) » Wed Jan 27, 2016 12:27 pm

Post by Psyche »

is there anything you guys think scum
usually
do that town
usually
don't?
youtube playlist extracter | donbot | game scraper | vca | setupsim | strategist | llm
User avatar
Psyche
Psyche
he/they
Survivor
User avatar
User avatar
Psyche
he/they
Survivor
Survivor
Posts: 10058
Joined: April 28, 2011
Pronoun: he/they

Post Post #7 (ISO) » Wed Jan 27, 2016 12:27 pm

Post by Psyche »

besides the obvious stuff!
youtube playlist extracter | donbot | game scraper | vca | setupsim | strategist | llm
User avatar
kuribo
kuribo
he/him
Fire and Brimstone
User avatar
User avatar
kuribo
he/him
Fire and Brimstone
Fire and Brimstone
Posts: 15463
Joined: August 21, 2007
Pronoun: he/him
Location: the beach, probably
Contact:

Post Post #8 (ISO) » Mon Feb 01, 2016 2:43 am

Post by kuribo »

A good one is the aborted bus. You gotta catch one of the scum first to incriminate the other but it goes like this:

Scum A has a wagon that is halfway and picking up steam
Scum B hops on for sweet town cred, maybe even helps with the case
Scum A's wagon fizzles for whatever reason.
OR A's wagon gets very close to lynch and suddenly Scum B isn't convinced.
Scum B hops off.
Scum A or B dies later
You recognize the signs and symptoms of B bussing A but not wanting them to die
You lynch the buddy
Mock the dead
Mock the living too
Join me on my quest to play every NES game! Some of them are awful.

Kuribo's read is foolproof: one night he was high on NyQuil, and he's ancestors reveiled Aureal's alignment to him. - Dessew
User avatar
GreyICE
GreyICE
Fifty Shades
User avatar
User avatar
GreyICE
Fifty Shades
Fifty Shades
Posts: 15455
Joined: December 15, 2010

Post Post #9 (ISO) » Wed Feb 03, 2016 6:11 pm

Post by GreyICE »

It is a fine exercise in astrology. When you are correct, astrology is great. When you are wrong, it is your fault for misinterpreting the pattern in the stars.

Kuribo is right in that you can tell much from how an individual behaved in one circumstance, but to divine the alignment of a large number reliably? Hah. And then there's the issue that as soon as it's public you have an intelligent entity deliberately fucking with it.

All my best VCAs are when I was scum and wanted a town lynch.

Which is not to say you can't learn a TREMENDOUS amount from a wagon. I once made a lynch pool of three people based on the fact that I didn't think scum would vote a certain way, and two of them were scum (in a game with 14 players alive). You can tell a lot by how people behaved on a specific wagon, but again this is quite a lot more involved
Show
That which is done out of love always takes place beyond good and evil


Official Visigoth

Read Mother Jones
User avatar
Titus
Titus
She/her/hers
Moon Walker
User avatar
User avatar
Titus
She/her/hers
Moon Walker
Moon Walker
Posts: 80249
Joined: May 3, 2013
Pronoun: She/her/hers
Contact:

Post Post #10 (ISO) » Wed Feb 03, 2016 7:50 pm

Post by Titus »

Voting Pattern analysis involves looking at the votes taking what is mod confirmed or cop confirmed and then looking at reactions from there. If you know what scum are likely to do in the aggregate, your VCA will be much more accurate.
Show
The scum had the misfortune of Titus being absurdly accurate on day one.Really quite impressed by that.~Drixx

You're letting Titus win the game by herself.Good luck now I guess.You have no chance to win.~Tywin

GTKTitus Part 2
Titus Academy

VLA Friday nights until Sunday morning.

All hail the Scum Empress!
User avatar
Antihero
Antihero
al;kdjfal;kj
User avatar
User avatar
Antihero
al;kdjfal;kj
al;kdjfal;kj
Posts: 15872
Joined: March 30, 2009

Post Post #11 (ISO) » Wed Feb 03, 2016 8:17 pm

Post by Antihero »

In post 10, Titus wrote:If you know what scum are likely to do in the aggregate

you won't, and attempts to do so will likely be you imposing some arbitrary and inappropriate generality
The distance between insanity and genius is measured only by success.
User avatar
kuribo
kuribo
he/him
Fire and Brimstone
User avatar
User avatar
kuribo
he/him
Fire and Brimstone
Fire and Brimstone
Posts: 15463
Joined: August 21, 2007
Pronoun: he/him
Location: the beach, probably
Contact:

Post Post #12 (ISO) » Wed Feb 03, 2016 9:58 pm

Post by kuribo »

In post 9, GreyICE wrote:All my best VCAs are when I was scum and wanted a town lynch.


I always think back to the first Gay Mafia, where DGB, DeltaWave, Katsuki, and I decided two things pre-game:

1) No bussing

2) We would consistently vote in a block because no one would expect the entire scumteam to be on every mislynch.

Worked great until DW got jailkept, then DGB and I counterclaimed the masons and Katsuki's PGO claim fell through

In post 9, GreyICE wrote:You can tell a lot by how people behaved on a specific wagon, but again this is quite a lot more involved


Yeah but after a point, you're not doing VCA anymore, you're just analyzing in-game behavior. I've found it more consistent to follow the votes and combine it with tone / attitude toward certain players I had a fair bit of accuracy in my heyday with statements like "If X is scum, Y and Z are his buddies. Y can be scum without X, but if Y is town, Z is probably town too." etc
Join me on my quest to play every NES game! Some of them are awful.

Kuribo's read is foolproof: one night he was high on NyQuil, and he's ancestors reveiled Aureal's alignment to him. - Dessew
User avatar
Claus
Claus
Mafia Scum
User avatar
User avatar
Claus
Mafia Scum
Mafia Scum
Posts: 1734
Joined: June 1, 2007
Location: Tsukuba
Contact:

Post Post #13 (ISO) » Wed Feb 03, 2016 10:08 pm

Post by Claus »

What data do you have? What format is it stored in?

I have some spare students who need a toy project, and I could ask them to feed your data to a neural network and see what happens over the winter break.

No promises, though :-)
http://www.youtube.com/watch?v=XVVmAG0RXmo
User avatar
AlwaysInnocent
AlwaysInnocent
Mafia Scum
User avatar
User avatar
AlwaysInnocent
Mafia Scum
Mafia Scum
Posts: 2453
Joined: November 18, 2015

Post Post #14 (ISO) » Fri Feb 05, 2016 7:44 am

Post by AlwaysInnocent »

In post 13, Claus wrote:What data do you have? What format is it stored in?

I have some spare students who need a toy project, and I could ask them to feed your data to a neural network and see what happens over the winter break.

No promises, though :-)
Damn it. People always have the same idea I have and then actually make it happen.
"You've been furthering the win condition of the Mafia even better than the Mafia." - Dierfire
User avatar
Claus
Claus
Mafia Scum
User avatar
User avatar
Claus
Mafia Scum
Mafia Scum
Posts: 1734
Joined: June 1, 2007
Location: Tsukuba
Contact:

Post Post #15 (ISO) » Sat Feb 06, 2016 3:40 pm

Post by Claus »

We can collaborate :-)
http://www.youtube.com/watch?v=XVVmAG0RXmo
User avatar
Psyche
Psyche
he/they
Survivor
User avatar
User avatar
Psyche
he/they
Survivor
Survivor
Posts: 10058
Joined: April 28, 2011
Pronoun: he/they

Post Post #16 (ISO) » Sat Feb 06, 2016 5:00 pm

Post by Psyche »

someone a couple of years ago attempted a machine learning approach to the issue
it didn't work out very well
youtube playlist extracter | donbot | game scraper | vca | setupsim | strategist | llm
User avatar
Claus
Claus
Mafia Scum
User avatar
User avatar
Claus
Mafia Scum
Mafia Scum
Posts: 1734
Joined: June 1, 2007
Location: Tsukuba
Contact:

Post Post #17 (ISO) » Sat Feb 06, 2016 5:44 pm

Post by Claus »

Aw. Do you have the thread, or know what kind of difficulties did they run into?

I actually heard of a professor working on machine learning for Mafia at a smaller university in my province, but I never thought about looking into it until now.
http://www.youtube.com/watch?v=XVVmAG0RXmo
User avatar
AlwaysInnocent
AlwaysInnocent
Mafia Scum
User avatar
User avatar
AlwaysInnocent
Mafia Scum
Mafia Scum
Posts: 2453
Joined: November 18, 2015

Post Post #18 (ISO) » Sat Feb 06, 2016 5:52 pm

Post by AlwaysInnocent »

Well, obviously machine learning isn't some kind of "magic trick" that solves any problem. Some problems are difficult to solve, but sometimes problems are incorrectly approached.

So I am interested to know how the problem was approached. Perhaps we could learn something from it.
"You've been furthering the win condition of the Mafia even better than the Mafia." - Dierfire
User avatar
AlwaysInnocent
AlwaysInnocent
Mafia Scum
User avatar
User avatar
AlwaysInnocent
Mafia Scum
Mafia Scum
Posts: 2453
Joined: November 18, 2015

Post Post #19 (ISO) » Sat Feb 06, 2016 5:53 pm

Post by AlwaysInnocent »

In post 15, Claus wrote:We can collaborate :-)
That sounds like fun!

(BTW. I am not really put off by people having similar ideas and actually making them happen. Quite the contrary!)
"You've been furthering the win condition of the Mafia even better than the Mafia." - Dierfire
User avatar
Performer
Performer
Mafia Scum
User avatar
User avatar
Performer
Mafia Scum
Mafia Scum
Posts: 4142
Joined: August 6, 2015
Location: California

Post Post #20 (ISO) » Wed Feb 24, 2016 9:02 am

Post by Performer »

Reviewing and self analyzing vote counts - I recommend this tool to any town player. Like others have stated though, by itself it's not a good-enough scumhunting tool. When combined with other measures, it can work wonders.
I’m an informed Miller who knows there isn’t any Loyal modifiers and there is a total of 4 scum.
-Morality
I'm easily the best person in the game at mechanics. I don't presume to be the best at anything else.
-Jingle
People tried
-RadiantCowbells
Post Reply

Return to “Mafia Discussion”