The problem with the GT1T school, to adopt the phrase, is that I think a lot of people mis-apply it, although maybe not GT1T. That is people getting lynched for having bad or illogical ideas when expressing those ideas is really pretty villagery for them, or at least the fact that the thinking is not logically valid is not a wolf tell

The problem with applying a random result theory to WW, especially one that looks at economic models, is that, unlike economics, WW is a binary situation. You either are or are not a wolf. In economics, there's a much larger grey area, and a specific companies situation changes.

That said, I have limited knowledge of economics, so if what I said is non-sensical, then I really don't care

1. The tone readers are right, the logic people are wrong. Quality of logic correlates with the person, not the role.

2. Confidence is great and necessary to lead a lynch, but claiming more confidence than you really should have is anti-village.

i'm going to go ahead and say that, logically, you're wrong

Logical is probably the wrong word, because it implies some correctness. However, it's the closest word I can think of to describe the concept wherein you show a thought process where you're working through the available information and coming to an internally consistent conclusion.

Wolfiness here is things like twisting facts to fit an agenda, ignoring inconvenient facts, or making logical leaps that they can't explain. Some villagers will do this, and they'll be incorrectly lynched, but imo, if you do these things as a villager that's a major leak in your game and you need to plug it, asap.

Of course, some villagers will also do things that lead to wolfy tone-reads as villagers, which is why there needs to be a synthesis of the micro and macro reads to determine the truth

It's not the quality of the logic, it's whether there's a thought process that's reasonable for that person to follow.

Notice there is no "Shortline School of Villaging"

I'm happy about that

Unless there is

Hmmmm

Zurvan
the random walk model applied to WW would basically assume each lynch was random (although you could easily adjust to assume 5% skill, or 10% skill, and thereby test what %age skill is needed for X to work or whatever) and then simulate the game. The only tough bit would be to factor in peeks, but i'm sure it wouldn't be very hard. The easiest thing to do would be then to run, say 1000 simulations under a given assumption of skill level and strategy (say a seer claiming to avoid a day1 lynch vs. risking getting lynched to freeroll) and test it against an alternative to see which works better. I'm no computer programmer, but this should be trivial (i could probably execute it in excel)

yeah I think I agree with you zurvan but I think a lot of the difference then becomes just that tone readers are doing a lot of that kind of analysis unconsciously, or just not posting it as Dustin says

I agree with the idea, I'd call it a tone read.

I'd put in the logic bucket stuff about voting history, "he'd never do that as a wolf", ...

This intrigues me. If you can give me some sort of step-by-step breakdown of how this is calculated, I can whip up a script.

Voting history & night kills are something else completely.

I use tone reads to clear villagers.

I use interactions to find wolves.

Interactions are pretty useful with inexperienced wolves but I think it becomes a lot more difficult when wolves get better and start thinking about how to fake them. Maybe I'm wrong but I don't feel like I left a lot of useful spew when I died as a wolf in any of my recent wolf games.

I don't use dead wolf interactions to find new wolves, I use dead wolf interactions to clear villagers.

I use alive people's interactions and their motives (who they are clearing / pushing) etc to find wolves.

ignoring the seer for the time being, to simulate a game you need a variable to contain the game state - (x,y) where x is the number of villagers and y the number of wolves. Then for a totally random game the lynch is a villager with probability x/(x+y) and a wolf y/(x+y). You rand() that, and have a nightkill (one fewer villager) which results in a new game state (x',y'). Repeat until you end up with y=0 or y>x. That's a game, with a result 0 or 1, depending on the win.

In Excel I would use 2^x*3^y for the game state so that I could hold it in a single cell, then a game would simply be a cell per day with the appropriate formula from on cell to the next, say in 4 columns or however many days there are. Then I could just drag down 1000 rows and simulate 1000 games, if that's clear.

To add skill to the game you would tweak the rand() so that the probabilities of a lynch would be biased towards a wolf getting lynched by a certain amount.

To factor in the seer would require a more complex game state, but it would be roughly the same and peeks/claims would impact the rand() probabilities for each lynch (i.e. 2 peeked villagers wouldn't be lynched so the rand() would be better, then the NK would reduce to 1 peeked villager)

Then ultimately you want to be able to tweak the rules for rand() and succession etc. to reflect different strategies.

Hope that's clear.