The Great Vote Count Analysis (Pre-Discussion)

[Psyche]

Let's say hypothetically that I have my hands on a dataset tracking every vote anyone's ever made across hundreds or potentially even thousands of completed non-theme games. I've got each vote's post number, target, and maker, whatever. I even know which Day they happened. And for each game in the data set, I also hypothetically have all the stuff you usually see in a mod's OP - stuff like who won, each slot's role/alliance/fate, replacements, etcetera. I also of course have every one of each game's posts.

What are some useful or interesting questions that I can ask of this data? I know that people examine vote counts a lot to figure out a game - what are some ways we can ground these sorts of analyses in evidence?

An organized list of exploratory questions posed as of the linked post

[Krazy]

% of scum self-votes to town self-votes
% of scum first vote busses vs scum first vote town votes

not sure if it's possible with your set but maybe % of scum L-1s that are not hammered

[Amrun]

% of scum hammers on town
% of scum hammers on scum
% of town hammers on town
% of town hammers on scum

[RadiantCowbells]

For each player X belonging to all of the set of games find out

What % of lynches with their vote on them as town flip scum
What % of lynches with their vote on them as town flip town
What % of lynches with their vote on them as scum flip scum
What % of lynches with their vote on them as scum flip town

[RadiantCowbells]

Can you determine what % of chronological time during a day phase someone was voting no one?

[skitter30]

- how often scum bus
- how often there's more than one scum on a wagon
- how often scum vote right next to each other

i really just want to make a bunch of (weighted ?) directed networks to look at voting behavior and see where that takes me

[Psyche]

can you elaborate on what that means

[skitter30]

can you elaborate on what that means

networks

I want to model games as different kinds of voting networks and see if any interesting patterns emerge.

And by 'voting network' i basically mean a network where a player would be a node and a vote from player A to player B would be a directed edge from A to B

I'm not exactly sure what I'd find but i kinda suspect that scum nodes would have different characteristics than town nodes.

And we can make lots of variations on this using different weights for the edges: how many times A voted B, for how many posts in a phase A was voting B, for what percentage of a phase A was voting B, etc.

I'm not sure how directly relevant any of this would be for a lot of the questions asked in this thread but it's something i've been wanting to look at for a while, and would be p doable with the hypothetical datset from the op

[gobbledygook]

- average frequency of town posts
- average frequency of scum posts
- average frequency of town votes
- average frequency of scum votes

[popsofctown]

How often do scum vote someone, vote somewhere else, then return to the original vote, and how often town do it in contrast

[RadiantCowbells]

i think divorcing whatever #s you're going to pull out of this from their actual contexts is going to significantly hinder your ability to draw conclusions that will help you reliably hit scum or clear slots

[Psyche]

well that's what the posts are for

[popsofctown]

i think divorcing whatever #s you're going to pull out of this from their actual contexts is going to significantly hinder your ability to draw conclusions that will help you reliably hit scum or clear slots

That's not really how statistics works. If 70% of people who selfvote in their first post are scum and your sample size is good enough, if you run around powerlynching everyone who selfvotes in their first post, you will hit scum 70% of the time.

If you evaluate context correctly, you're only going to improve on the values the stats gave you. If you don't lynch the first-post self-voters that selfvote in response to a joke fake mechguilty because that's a mitigating factor then you can go up to hitting scum 90% of the time that you follow through on lynching a first-post self-voter. But failing to do so won't "hinder you" below that base 70%, it's instead that you've left some value on the table (and none of these is likely to be as good as 70% so you can't actually afford to leave value on the table).

Of course that "you. will. get. seventy. percent." depends on you dogmatically applying the stats to every single situation they are applicable. If you only sometimes resort to the stats inferences and specifically only do so when you haven't gotten what you need out of conventional dayplay, you've created a selection bias, and that bias likely makes the stat inferences more valid or less valid in the subset of games where you involved them. My money would be on the stat inferences being less valid in the subset of games where you "needed" them.

I suspect your viewpoint isn't actually very different from mine and this is a glass half full thing but I'm not sure.

[Amrun]

- how often scum bus
- how often there's more than one scum on a wagon
- how often scum vote right next to each other

i really just want to make a bunch of (weighted ?) directed networks to look at voting behavior and see where that takes me

Yes to all of these

But I’d split the latter two into town and scum wagons

[RadiantCowbells]

I think purely off of votes you're never gonna get any evidence that gives you better than 50%ish chance of hitting scum
I could be wrong @popsofctown but I simply don't think the effectiveness numbers are gonna be high enough for it to be meaningful

[popsofctown]

RC the secret to enjoying the stats is to be so bad at mafia that the weak statistical correlations are more accurate than your sorting from analyzing posts.

Come get on my level.

[Psyche]

we really don't just have the votes, though
i've been learning a lot of text analysis techniques from my research; we can design analyses that are sensitive to context if we come up with the right questions

(also 50% is pretty good)

[yessiree]

50% accuracy is really not tho, you can flip a coin and get basically the same results

also ego

[Psyche]

Relative to a base rate of like .28 for a 2/7 setup, for example, .5 would be great from just looking at votes.

But i don't really think collecting a bunch of statistics comparing town and scum rates of some behavior in a given situation is going to help anyone find scum in a real game. Even if one of these comparisons found some substantial factional difference on average, I dunno how'd that information would be applied in particular situations.

Think I will do these exploratory analyses, but am hoping I can try building and testing models of voting behavior that apply across situations. Maybe I'll start with a naive model that assumes every player randomly votes for whomever they don't know is in their faction, see where the data diverges from that, and go from there.

[yessiree]

I believe you can still get some meaningful patterns. I think skitter's idea of representing the votes in a graphical model sounds very promising, I wonder if we can finally do some quantitative analysis on busing (I have some ideas on how it could be implemented in networkx https://networkx.github.io/)

[N]

guys psyche isn't going to do any of these things. he just likes making extravagant promises he won't keep

[Psyche]

wanna do an avatar bet about it? a one month avatar if in one month's time i dont public a dataset of several hundred games, half the proposed analyses posted so far, and some basic modeling?

[chamber]

Taking that bet sounds like a win-win to me.

[skitter30]

I believe you can still get some meaningful patterns. I think skitter's idea of representing the votes in a graphical model sounds very promising, I wonder if we can finally do some quantitative analysis on busing (I have some ideas on how it could be implemented in networkx https://networkx.github.io/)

Yeah, i think that this is a natural and intuitive way to model games, and would probably lead to some interesting conclusions

The key i think would be to figure out the best way to weigh edges

I dont think it would be too hard to implement with a dataset consisting of all votes cast; that's basically a directed edge list and from there it would be p easy to feed it into some network analysis packages

[Krazy]

% of scum self-votes to town self-votes
% of scum first vote busses vs scum first vote town votes

oh yeah and a version of this that excludes self-hammers