The essence of epicmafia18
Epicmafia is essentially a game about which player is randomly assigned the clear role, and which players they randomly choose to vote out.
If the player that gets lynched is town, the clear says "it's not my fault they were so scummy!".
If the player that gets lynched is mafia, the clear says "they were obviously scum, I'm so good at reading people!"
After their choice, they take a mental note; In the future, they will be less or more likely to scumread people that act like the voted out person depending on the true role of that person.
If the clear made a mistake, and someone in the graveyard had a different read, i.e., made different mental notes throughout their time playing, then the person will become angry at the clear and say "they were obviously town!".
You think people give away which alignment they are through what they type. You think figuring it out is a complex process. You think there are hundreds of variables to take into account.
You think with infinite wisdom, you will be able to account for all of the variables made available to you to be able to reliably read people. You think your ability to read people will improve over time as you experience more and more games.
This claim is unfalsifiable. You would need a person with infinite wisdom to test it. But you accept this claim without questioning it.
One test we can do is to look at the win ratios of the people that should be closer to infinite wisdom than anyone else. These are the people at the top of the total-score board. SuperNova, Sinek8, MafiaGod, Diabolik, and grifiest have an average win-rate of around 51%; the highest any of them have is 52%. Cumulatively, these accounts have played approximately 11,300 games.
The people that have years of experience cannot achieve winrates significantly higher than anyone else on the site that knows how to play and tries to win.
This indicates that people to not improve at mafia over time once they reach a very low skill ceiling.
The "scumtells" and "towntells" in your mental notebook are just the product of not having enough data to realize that no significant correlation can be drawn between the things people type and which role they truly are.
You should not delude yourself into thinking that you will improve at mafia over time.
Some players do achieve significantly high winrates (55-60% across large numbers of games). These are the players that usually play with their friends, or make a conscious effort to be very friendly to everyone. These players are less likely to be "randomly" chosen to be voted out when the clear has to hammer.
I can assure you un-cheated high % accounts aren't all the way you describe; they distribute from very agreeable to completely disagreeable people. Some people are very good intimidating others to do what they want, and that's the least bit "very friendly to everyone".
As for what the strategies are effective, there actually is a lot of research on this topic, one that makes a strong significant statistical difference is keeping an accurate log of events, and who said what when in response to what; which EpicMafia does for you automatically.
However, the basic skill sets which require deductive reasoning as well as the ability to persuade and use rhetorical techniques to convince others to your way of thinking is an actual subject of study. Those who are better at persuasion, logic, and deductive reasoning are going to perform better than those who aren't.
Mafia isn't a "perfect" game in the theoretical sense, it's one that specializes in incomplete and imperfect information, and as such there is a Nash equilibrium that will provide you with the best results.
According to David Sklansky, decision making in the face of incomplete information can be learned and in some sense, perfected, at least in the incomplete game known as poker, and in Mafia it's no different.
Each decision you make in Mafia is based on a projected payout of each option according to the parameter of all possible games. The goal isn't to win every single game, but extract more value from each game in order to win the most possible games. Of course you aren't going to win every single CC or lynch mafia every single time, but the strategies overall will grant you the highest chance of winning you possibly can given the conditions present in the game.
Vanity you nerd, you're overlooking the actual game with all your meaningless big words. One thing you're overlooking is the mechanics & the math - monkeys with RNG playing coco's will approach 55% winrate, while in fancypants they will approach 48%. You can't talk about convincing in the context of fitness without mentioning how the actual game works. Another, is the meta. Pointfarms are a perfect game that requires no log-learning or anything of that sort.
As for the regression fitness, yeah, I imagine that certain setups are going to be more imbalanced than others, and have more to do with random dumb luck then mechanics, but generally in the comp rotation as you know, those types of games are generally filtered out.
Perhaps I'm misunderstanding your critique of my post, I'm generally trying to address the OPs main point that incomplete and imperfect games don't require skill, which I don't think that's necessarily true. If I'm wrong about his general thesis then I guess I'd be wrong about my post. But I don't think that's the case Linker.
The reason that is wrong is not because its an imperfect game, but because of how he thinks improvement is a linear that starts from the 50% and all of these people are just trying their best.
Putting how half the names he mentioned throw on a regular basis aside, these people didn't start from 50% because the setups they used to play were 45%-ish. getting 52% on a 45% setup that has situations which are forced could possibly mean the person wins out 66% of their CC's, which is incredible.
Not only this, but some people are better than others. He says that's a rare edge case of really agreeable people, and I say that has little to do with how agreeable or experienced you are and much more meta, and only here is where your answer comes in - skill, which you explained well.
Look at the history of poker players; are the players who were at the top of the game in 2009 the same as they are now? Do the players who win the WSOP main event exclusively represent the strongest active players at any given time?
Mafia (like poker) isn't so much a game with a low-skill-ceiling but a game where the number of variables a player can reliably control is relatively small compared to the total number of variables in the game.
If you can understand that concept, then perhaps you will be able to interpret the statistics you brought up properly.
There is an informed minority, who doesn't know the roles of the townies besides "town" and an uninformed majority, which the only information that's revealed after the fact like in a lynch.
The outcome is not "random" nor are all the choices players choose like whom to lynch/kill. Most of the game revolves around getting information and processing it.
Familiarize yourself with incomplete and imperfect games, and generally game theory if you want to understand where the error in your reasoning is, because if this is the first time you're encountering this subject you really aren't getting a complete picture.
Everyone has their own methods to read people. But no one's methods can be coherent with anyone else's. As soon as someone decides a certain action indicates that someone is mafia, they will stop doing as mafia. On a macro scale, this phenomenon will mean any tells stop meaning anything. Everything gets flattened so nothing is consistent. In the game I linked, people accuse other people of being mafia for nonsense reasons, or reasons that they can't seem to begin to articulate. They act like they are sure of their reads, but they have just as much accuracy as they would have had if they had guessed. People think they are ahead of the curve regarding what actions are scumtells; if that were the case then someone on the website would be able to read people consistently, but no one does. A person trying to make reads will inevitably get lost in a cycle where they question whether an action is a towntell because a town would do it, or whether it's not a towntell because a mafia would fake it. Or they will question whether an action is a scumtell because a mafia would do it, or whether it's not a scumtell because a mafia would avoid doing it. If their choice was correct then they will assume they are improving, if it was wrong then they will assume they have room to improve. But in reality, they are trying to get good at predicting the result of a coin flip.