Behavioral Analysis, Chapter 2 – Evaluating the Strength of Evidence

Behavioral Analysis, Chapter 2 – Evaluating the Strength of Evidence
By Azrael

So, you’ve got a case on a player you think is scum. But how can you tell if you’ve got a rock-solid case, or a crumbling foundation?

There’s a good deal of misinformation floating around about the relative strength of scum tells. It’s certainly not a cut-and-dried field. However, there are a number of tells that seem to have proved their worth over the years, just as there’s a body of tells that have proven shaky at best, and baseless at worst.

For my own purposes, I use four classifications of scum tells to help me evaluate their strength. Universal tells, logic tells, possible tells, and emotional or attitude tells.


Universal tells are the traditional slip-ups that have been identified over years of playing and study. They are activities where the most common explanation of the tell is that the player is scum, and there is a reliable, proven correlation between the behavior and mafia players. A short list of some of the more common examples could include:

-Malicious OMGUS attacks in response to pressure.

-Overaggression in voting patterns.

-Atypical patterns of participation, such as posting only enough to avoid scrutiny, or avoiding certain critical discussions within the town and commenting only on tangential issues.

-Waffling and taking both sides of an issue.


-Poor, unpersuasive, or faked vote rationales.

-Unusual timing of a claim, or role information.

Oftentimes, these tells can be difficult to spot if you’re not paying close attention to a player’s exact wording.

For example, consider this post, ripped from Chapter One:

I didn't like your vote because I felt it was too early and was disrupting speculation on the actual game mechanics,
which is something I want to know. You do seem town to me though.

The last sentence of the post is a clear example of scum-like waffling, taking both sides of an issue. But does that mean that anyone who waffles or examines an issue from two different sides has committed a scum tell? Absolutely not. So how do we tell the difference?

I would suggest that we ought to expand on the causal analysis techniques described in chapter one. What is the cause of this player’s waffling? Is it because it’s a really tough call and the evidence is conflicting? Or is it because the player is mafia, and can’t decide what a townie would actually think? If we can guess the true cause, we can guess the alignment.

With most of the universal tells, it’s easy to see how a scum mindset can be the cause of these behaviors. For example, scum players often fake their vote rationales because they’re inventing them on the spot rather than arriving at them through careful consideration. Through paying close attention to their wording, you should be able to gain a better idea whether players are manufacturing a flimsy justification, or stating what they truly believe.

But be careful. Just because a player commits an action that is considered a universal tell, don’t assume that the scum rationale is the cause. Consider also the possible pro-town causes of the player’s actions. Oftentimes, the pro-town rationales for committing a universal tell can be equally strong or more strong than the scum rationales. In those situations, the behavior you’re looking at should not be counted against them.


Logic tells are errors of argumentation, such as evasion, misrepresentation, OMGUS, and ad hominem. While equally well known to most players as the universal tells, many of us fail to distinguish between the two. Most often, we make the mistake of equating logic tells with universal tells. Sadly, this is not the case, and this failure to acknowledge the differences between the two has led to countless mislynches over the years.

The reason this occurs is that logic tells are frequently caused not by alignment, but by idiosyncrasies of the player. For instance, ManbearPig in one of the examples from Chapter one refused to respond to further attacks after answering them several times, what many players would call a clear-cut evasion tell. But the reason behind the “tell” was that he felt that he had replied to them sufficiently and at length already. While a scum might feel similarly, the existence of a very reasonable townie rationale weakens the argument that his refusual to respond indicated a scum alignment.

It is critically important that townies recognize that the negative effect of an activity, such as ad hom or misrepresentation, is not the best indicator of whether it is a genuine scum tell. The best indicator of whether a tell is genuine is if there is not a more reasonable and more likely townie explanation for the action. For instance, MBP was an inexperienced player, and he was signaling frustration with the debate throughout his posts. Many players ignored these subtle signs and instead assumed that the reason for his evasion was that he was attempting to hide from exposure. However, there was no concrete reason to believe that he was trying to hide as opposed to simply being inexperienced and frustrated. The context of the activity actually supported the townie rationale most strongly.

Part of the problem with logic tells is their tendency to provoke emotions. When tensions are running high, it becomes easier to begin to see your opponent as an enemy. This, unfortunately, often leads townies to take shortcuts in their arguments. Then, one side may begin to suspect that their opponent isn’t arguing in good faith (and perhaps they are not), at which point they begin to assume that the logic tells are evidence of their opponent’s guilt.

While ad hominem, evasion, and straw-manning are frustrating to encounter, they usually have little or nothing to do with player’s alignment. They can sometimes be indicative of a scum mentality, because there is a weak correlation between sloppy play and scum alignment. But more often than not you will find the weaker, less seasoned members of the town committing these types of activities regardless of their alignment. When these types of tells occur, it is critically important to read between the lines dispassionately. Is the misrepresentation intentional? Is the evasion a function of time, or frustration, or stubbornness, or is it out of fear? Is the ad hom a calculated device, or just pure emotion?

Logic tells should be used to support cases only in rare and exceptional circumstances. They will mislead you more often than they will help you.


The next category, possible tells, differs in one significant way from universal tells. Similarly to logic tells, they rely on assumptions. If such and such is the case, then that player may be committing some universal tell, such as defending a scum buddy for poor reasons. There is an additional logical step involved in believing that it is a true scum tell: we must believe in condition X, in order to think that Y is a scum tell. These assumptions more often than not prove to be false rather than true.

For instance, in the game with Pale Mage, a number of players involved with his lynch pointed to the difference between one of his short, acerbic replies and a longer, more nuanced explanation he offered later, as a sign of his being coached by his fellow mafia players. IF their assumption that he was coached was true, then it would be a scum tell. But if their assumption was false, or ungrounded, then it would be not be a tell at all. Once again, it is absolutely critical that the basis for the assumption is carefully examined. The change in his posting style could easily have come from any other number of variables, and in time that proved to be the case. But because they made that paranoid assumption, something that should never have been considered a scum tell to begin with was used to create a senseless mislynch.

As a general rule, conspiracy theories and complicated alleged gambits rarely turn out to be true. In most cases, what the activity seems to be on its face is exactly what it is in truth. Paranoia is one of the gravest and most common dangers to the town keeping its judgment objective and insightful. Instead, players should concentrate on what is most likely to be true, and not on far-fetched worst case scenarios.


Emotional, Mindset, or Attitude Tells

Unlike the other tells, mindset tells aren’t based on a player’s conduct, but on the more intangible behavioral traits lying behind their posts. Despite being more subjective, they can often be quite useful in determining their alignment, because players tend to exercise less control over them. Just as certain universal tells are usually caused by a scum mindset, the emotions and attitudes players show during the game can be very useful clues to their alignment.

A typical townie mindset could be described as inquisitive, pro-active, confident and decisive, self-righteous under pressure, and unafraid.

Some of the more typical scum fallacies are self-consciousness, contemptuous dismissal of others’ arguments, posting just enough to avoid notice, deceit, manipulation, being prone to faulty logic, showing more concern over plans and irrelevant debates than with finding scum, a tendency to use uncertain language in connection with lynch targets so as not be proved wrong later, a tendency to attack weaker or more inexperienced players, and being nervous or overbold in presenting their role-claim.

Just as with universal tells, however, it is important to consider the context that an emotion or attitude is in. Take Manbearpig’s refusal to continue an argument, for example. While refusing to continue an argument is an extremely anti-town attitude, he had a fairly decent explanation for it. To sum up, mindset tells and universal tells are the type of evidence you most often want to ground your cases on. They still need to be evaluated to make sure that they’re legitimate given the context, but if you can fit those facts and subtle clues into a clear and reasonable picture of what was going on in your suspect’s mind as scum, and show that the townie picture makes no sense, then you have the hallmarks of a very promising case.

In looking for mindset/emotional tells, bear in mind that these tells are often your best way of identifying and clearing fellow townies from consideration. The focus of most scum players is to avoid committing scum tells. Only the very best, most experienced scum are knowledgeable enough, skilled enough, and dedicated enough to consistently seed authentic-seeming false townie positives in their posting - less than one percent of players. As a result, analysts who can are able to detect genuine townie mindsets in other players' posts can be invaluable assets to the town, by clearing wide swathes of the town from consideration and stopping mislynches before they occur. In addition, once you have enough confidence in your abilities, you will be able to use this method to narrow your search of candidates by process of elimination, and lynch those players who have failed to provide any confirmatory town tells. In many games, it may be easier to find the scum by figuring out who the townies are, than by seeking out the scum directly.


1. Universal tells can be strong evidence of guilt. However, it is vitally important to examine the context to see if a reasonable, pro-town rationale exists.

2. Logic tells and possible tells are unreliable and should be used only with extreme caution.

3. Analyzing a player’s emotions and attitude for scum or town patterns can be a good indicator of their alignment.

4. Be alert for townie emotional and minset tells in the players you are analyzing, as this is your best chance to avoid a mislynch and clear players from consideration.