You are viewing the Wiki. To play the game, visit the forum.

Mastina's VCA Guide

From MafiaWiki
Jump to: navigation, search


Original Publication: March 09, 2017 by Mastina.

Last Revised: March 23, 2017 by Mastina.

Original Thread.


So I've personally observed that when it comes to Vote Count Analysis, most players are pretty shitty at doing it. I've also recently been receiving a lot of requests for me to share with others the exact process behind my own method. So I decided I'd finally break the drought in Mastin MD and write this. I apologize it's not succinct; the formatting of the advice is a WIP, however, the content is not.

Let's start with the basics.

VCA Defined:

VCA is a tool using votes, analyzing the game to help form conclusions.

Most importantly to note in that definition...

VCA is a tool.

It is not the end-all, be-all of scumhunting. It is not something meant to be the entire basis of reads; it is best utilized to help augment evidence and highlight trends you might otherwise miss. Never rely exclusively on VCA, morphing reads to fit the VCA. While that's the most critical rule, let's establish some more, for good VCA.

Objective vs Subjective

Present objectively; use subjectively.

It's important to not confuse the two. When you do VCA, it is hard to deny the votes happened. You are presenting objective facts. However, your interpretation of those facts is your opinion--thus, subjective. You should always keep this in mind when you write your VCA. You are giving a narrative of things as you see it.

Never be arbitrary

The best VCA is VCA which does not blindly make arbitrary calls: "scum MUST be on this wagon", "scum CAN'T be on this wagon", "this player MUST be town", all commonplace as a result of poor VCA methodology. Relevant to this,
Scum going down have a natural tendency to "suspect" a buddy. That's not a hard tendency and you never want to over-value it, but the idea of TRing someone just because a scum whose vote didn't matter stayed on the slot is simply not good.

When making good VCA, you do not assume scum act in some predefined way--you pay attention to what scum are likely to do in the given situation.

Best without bias

Don't color in names of unflipped players unless you have mechanically good reasons to do so. (E.g. masons, cops, cop investigations. Those would all be decent reasons to call an unflipped player town.)

For instance, there's a difference between "player generally thought to be town", "player with STRONG reason to be considered town", and "player with reason to be called confirmed town". Under no circumstances do you color personal townreads/players generally thought town. If you decide to color unflipped players who have a strong reason to be considered town, you probably want to color them differently from those that are absolutely confirmed as town.

Basically, you want to avoid having preexisting reads bias your votecounts.

VCA has different purposes

Always have a clear idea of what your VCA is actually meant to do: is it meant for you to find initial reads? Is it meant to call into question your established reads? (And then either strengthen/weaken them.) Is it being used to make a case? All of those are still forms of VCA, but they all have different times they are used and different reasons to be used.

And relatedly (and critically),

VCA has different levels

These different levels vary from person to person, but basically, there's a difference in thoroughness in how a player uses VCA, yet these different levels of VCA are no more nor less valid than the others, simply...more in-depth than others. For the sake of this thread, I'll share my personal levels, as I have broken them down.

Situational VCA

Also known as a "gut feels" approach to wagon analysis--I'll look at votecounts, and intuitively 'feel' what each wagon makes me feel like and why. "I like this wagon". "I hate this wagon". "These players on this wagon feel good". "These votes on this wagon feel bad". Overlapping those four things is a basic form of VCA, where I am instinctively feeling out what SEEMS right.

Usually, it's used only in the earlygame (especially D1), though it can be used later when observing forming wagons. It's little more than guesses layered on top of guesses, though because it requires little thought/effort, it's a nice lazy method of VCA.

Basic VCA

Usually done after D1 when there are players flipped. This is thrown together votecounts, and from the votes, forming predictions. Basic VCA works by extrapolating facts from some base assumptions, to run basic probabilities. It's still highly subjective, not so much a weapon as an augmentation for what's already around: pointing out little things here and there which help.

Pattern Mapping

The advanced form of VCA, a requirement for it functioning properly is days' worth of data, preferably with the full rainbow of flips a game has to offer. Pattern Mapping is looking at flips, observing trends, noting how and why wagons formed, to find patterns in the data and providing interpretations of these patterns. It's a combination of all that VCA has to offer--not only making predictions, but also using judgment. This pattern recognition is what VCA is meant for. It is analysis of votes (the core of the town) to find information and answer questions.

With those rules established, here are a few additional guidelines I use. These tend to crop up especially in the higher levels of VCA I do. While I may not run every one of these through my head when doing situational VCA, when doing pattern mapping, these are quite literally prerequisites.

Wagon Composition

How many town/scum are on/off the wagon?

You'll note a rule was not making arbitrary assumptions--so don't make a statement of "this many players must be scum/town" on/off the wagon. Instead, ASK "how many are scum and how many are town?". The best VCA is fundamentally asking questions and seeking answers to those questions. Your aim with VCA is to find where scum are versus where town are...not to have a predefined notion of that already.

Scum lynch:

Look for bussing.

The first place I look for scum on a scum lynch is on the wagon (ESPECIALLY a D1 scum lynch--it is ridiculously hard to lynch scum D1 without the assistance of scum because of town chaos and paranoia). While there is obviously no guarantee that there actually is scum there, I put analysis in to determine where the scum would be if not on the scum wagon. (More on that below.)

Town lynch:

Look off the mislynch wagon.

The first place I look for scum on a mislynch is OFF the wagon--I try to determine how many scum (if any) there are, and who the probable candidates are. Sometimes, there are situations where there are none to be found, or far fewer to be found than expected...and in those situations, by virtue of doing this analysis, I have a better idea of how good or bad the mislynch wagon was.

Regardless: counterwagons

I look for the presence of a counterwagon, or far more tellingly, the utter lack of one. If there is no discernible counterwagon on a scum lynch, it's a safe bet scum bussed. Maybe even heavily. If on a scum lynch there is a counterwagon on a player who is known to be town, then the names there have a decent chance of being scum. This is not that common anymore, but sometimes it does still happen. If on a town lynch there is a counterwagon on a player who is known to be scum...then there's a reasonable chance you can use the safe/standard model of "scum voted town, town voted scum" with an above average accuracy.
that guy sitting out the day's lynch when it's town v town? That's scum with substantial probability.

If on a town lynch there is a counterwagon on a town player, then the scum didn't care which got lynched--the scum are likely to be about evenly distributed between each wagon, and if there's players on neither wagon, there's a disproportionately high chance that the players off both wagons are scum.

Frequent vote overlaps

I look for names who are repeatedly voting together on multiple wagons--this is a fairly decent sign that there's probably scum in there. While there are obvious exceptions (a townbloc which moves votes as one; masons), usually this is a fairly reliable indicator that at least one of the players involved is scum. You need to figure out if it's one or all of them, and if just one, WHICH one.

On that note,

Breaks in overlaps

When DON'T those players vote together? This can be a critical hint as to the alignment of the voters. Pay close attention to who the vote-buddies have diverging votes on. This can help you read the vote-buddies and also illuminate interesting facts about the gamestate.

Data Patterns

You're looking for the basic trends in data.

That is, you are aiming to see common, frequently occurring patterns. These overlaps, these recurrent themes, tend to be where the value of VCA is at its greatest. Hard data. Data which raises questions about how the data came to become true, begging for answers (which are yours to give).

Observe the gamestate

To steal from mhsmith0 again:
day 1 scum lynch, days 2-4 town lynch = scum lynched was bussed, probably by both buddies (if he hadn't been bussed, then why did town suddenly become so ineffective after a day 1 scum lynch?)
More generally, the gamestate of the game makes a difference. What made the lynches on scum be lynches on scum? What made the lynches on town be lynches on town? You want to know this, because the answer tells you who the scum are.

Understand a push

not surprisingly, the people lining up pitchforks the hardest were town, while scum were setting up future pushes / staying out of the way. It's hard to replicate villagery fury at someone who really fucked up what they were up to as scum.
Similar to observing the gamestate, you need to know the emotions behind a lynch, and the emotions behind votes NOT on a lynch. In particular, you're looking for force of the push. If the player's push did nothing, then...
Scum could EASILY vote park on a "push" going nowhere. The question is if the wagon ever did anything. If not, congrats, it's >rand scum voter.

...It's a fair bet you've found scum.

There are a couple of other quirks I use in my VCA.

Combine Wagons

Of identical size, at the very least.

All (1) voters as a wagon, all (2) voters as a wagon, all (3) voters as a wagon, and so forth. The idea behind doing this is to counteract a common scum strategy--vote dispersion. Scum very frequently divide themselves up evenly, and if you combine identically-sized wagons into a few larger wagons, you can get a much better idea of probable scum distribution. (I do this mostly as a guideline.)

Count Not Voting

Not Voting is, itself, a wagon.

...Past a certain point, of course. If a player hasn't posted yet, then obviously they couldn't have voted yet. But if the player has posted yet is in Not Voting, they have made the active choice to not cast their vote. As a result, Not Voting is a wagon in of itself.

Overlap the two above for the best results. However large the Not Voting wagon is, combine it with an identically-sized wagon to get an idea of where scum are likely to be hiding. This is most effective when you have data on the players who were wagoned, and is also not a hardfast rule. (I don't treat combined wagons/Not Voting as a holy gospel of gold. It is difficult for scum to counter--not impossible.)


Pay attention to "invisible" wagons.

That is, make note of wagons which rose, then fell, between votecounts. Mods are usually good about getting in the necessary votecounts for effective VCA...but what if there were ten votes on a page and then ten unvotes on a page for a single player? (Hey, it can happen!) That would be noteworthy on the VCA but a mod who pagetops their VCs might entirely miss the wagon.

So, if you can, manually track these wagons which came and collapsed quickly--note all of the names on them. These extra wagons aren't inherently on an alignment, and they aren't inherently driven by one alignment. However, they are contextually key pieces of the game, and their absence from analysis will taint your results.

(Incidentally, this is why I do the extras in my votecounts. The things I track automatically keep those in mind, so that players can see them.)


Circumstances are everything.

In particular...were scum interested in saving the life of the scum lynched? How was a player off of the lynch acting? Frustrated, disinterested, detached, what were they doing? Fighting hard or sitting idly by?

What is it that was going on to generate the wagons as they were? For instance,
Generically you MIGHT suspect the off-wagon or non-voters more than the on wagon voters, but the only real question you need to ask was "was this lynch wagon villagery". Because if it felt villagery and righteous, you lynch off wagon, and if it was bullshit, you lynch on wagon.

...Lynch from a wagon driven by scum, and don't lynch from a wagon driven by town.

Always, keep in mind what each wagon represents, keep in mind what each player is doing, why they are doing it...think of what is the overall picture behind the wagons. This context behind actions is where you'll find the full strength of VCA.

Final Notes:

I tend to not do all of this, all at once. Like most of my advice, this tends to be "do as much as you can, as often as you can". The more thorough, the better. You should always keep in mind it is possible to beat even the best VCA. The best advice I can give is that you have to trust your instincts when it comes to what in your VCA is accurate and what in it is not, so long as this hunch of yours is not confirmation bias warping the evidence to the read. Look for the things which make the most sense, and trust in that. As long as you're avoiding shitty arbitrary assumptions of things which "MUST" be true, you're probably going to be right.