Site Meta Thread

[Not_Mafia]

Quick at what?

[mhsmith0]

Actually I’m quick

[Not_Mafia]

I'm a bisquick stick

[Psyche]

i cope every day with overwhelming insecurity

[LicketyQuickety]

i cope every day with overwhelming insecurity

As do I only with anxiety instead of insecurity.

#themoreyouknow

[implosion]

Motivation reading is definitely a specific tool, and it's probably the tool I use the most. It's also a way of looking at the game in general, but as a tool (at least the way I use it) it's basically a way of looking at a statement or a post or an action, considering each possible alignment's potential motivation for making that post, and then considering which of those motivations is most likely to have existed in the first place. A similar process can be used in thinking about the line of reasoning that that person would have had to have in order to make that post. It's sort of Bayesian in nature; the probability that someone is a given alignment given that they made a given post should be correlated with the probability that they would make that post given that they are that alignment. It's useful, especially in conjunction with some light meta, for getting reads off of things that a person may not have had any motivation to say as one alignment, or that may have required a thought process to happen as one alignment that is just unlikely.

It's not a be-all end-all as mastina says but it's pretty close.

Also in the interest of stating a likely unpopular opinion: I think that meta reads off of a single game can be

extremely

powerful in the right situation if you are very careful with them. Particularly when the game you're meta-ing off of was a recent scumgame and where you think the player in question is town in the game you're playing.

[LicketyQuickety]

Motivation reading is definitely a specific tool, and it's probably the tool I use the most. It's also a way of looking at the game in general, but as a tool (at least the way I use it) it's basically a way of looking at a statement or a post or an action, considering each possible alignment's potential motivation for making that post, and then considering which of those motivations is most likely to have existed in the first place. A similar process can be used in thinking about the line of reasoning that that person would have had to have in order to make that post. It's sort of Bayesian in nature; the probability that someone is a given alignment given that they made a given post should be correlated with the probability that they would make that post given that they are that alignment. It's useful, especially in conjunction with some light meta, for getting reads off of things that a person may not have had any motivation to say as one alignment, or that may have required a thought process to happen as one alignment that is just unlikely.

It's not a be-all end-all as mastina says but it's pretty close.

Also in the interest of stating a likely unpopular opinion: I think that meta reads off of a single game can be

extremely

powerful in the right situation if you are very careful with them. Particularly when the game you're meta-ing off of was a recent scumgame and where you think the player in question is town in the game you're playing.

I agree with most of this 100%. Completely correct about the line of reasoning and whether it comes from Town or Scum. This is often what I try and push in a Mafia game: I look for things that don't makes sense based on what I know of that players ability (which if I am playing with them for the first time, this is an ongoing process until a certain point and I can conclude what their playing ability is), what their assertion is, what their reasoning for their assertion is, and whether it is of sound logic. I have noticed that Scum more often make mistakes in reasoning much more often than Town. This is not the same thing as motive. There are times when Town do stupid things like vote themselves. That's where knowing your player comes in. If you know them to be a loose cannon, then them voting themselves as Town makes sense - especially if they are a new(er) player.

So I think that using meta has different uses from person to person. Some people just use it as a guid to tell how smart the person they are dealing with is, while other people look for very specific things within that players meta that can be a high confidence read based on the specificity of what they are looking at. There is ofc a lot of grey area in this and some people mix and match or use a bit of both.

[implosion]

Yeah basically entirely agreed.

There are times when Town do stupid things like vote themselves.

I think the technique is especially powerful in those kinds of situations. This is sort of my go-to example from many years ago where I got a really strong townread on someone for selfvoting on page one because of the phrasing of the vote and how he reacted to people's reactions. There's also an element of tonereading here that made the read stronger.

Generally anyone who says "X is a town/scum tell" where X is something simple like selfvoting, or being logical, or tunneling, or even quickhammering, is drastically undervaluing context, and motivation reading is really good at figuring out whether a given action in a given context is a good tell.

And yeah agreed about meta. It's useful sometimes to fully check someone's meta and see if they're capable of faking something as scum or if their games feel a particular way or have particular qualities as particular alignments, but it can also be useful to use it at as simple of a level of player X is really inexperienced and acting like it so their townslip is probably real (which I also have an internal go-to example of).

[Ellibereth]

how often does doing this motivation stuff get you a wrong townread.

[LicketyQuickety]

how often does doing this motivation stuff get you a wrong townread.

IDK who you are asking, but generally my Town reads are pretty good but my Scum reads are not very good.

[Ellibereth]

whats "pretty good" for town reads.

like 77% accuracy is just around random.

[LicketyQuickety]

whats "pretty good" for town reads.

like 77% accuracy is just around random.

I usually do tiers.

Tier one is generally much more accurate than not, even D1. Generally there are only like 2-3 people there tho.
Tier two is right more often than not.
Usually you find the Scum in my Null and lower lists.

Ex: viewtopic.php?p=9689644#p9689644

[Ellibereth]

you should actually measure the "more often than not" to see how much better than random you are .

I think most people are kinda vague and just judge their read accuracy by feel after games.

[implosion]

tbh i haven't played enough recent games (and my two most recent were scumgames) to have a sense strongly in mind of how accurate it is in any numbers. I'm sure I've gotten incorrect strong townreads with it but none come to mind because i'm biased and only remember the times it was right~

[LicketyQuickety]

you should actually measure the "more often than not" to see how much better than random you are .

I think most people are kinda vague and just judge their read accuracy by feel after games.

I suppose I can start doing that since I have new account/username/wiki now.

[TwoInAMillion]

We should have a Win Above Replacement rating similar to the baseball stat that calculates how good of a player you are.

[Ellibereth]

We could also start a rating system for each alignment.
Using something like msoft's true skill.
Can x10 or x100 it to make the numbers prettier.

Everyone starts at same amount now and after every game just take avg rating of each team before game with the result and do + teamratingchange/teamsize to everyone.

and just not count multiball and bastard and 3p games and stuff like that for simplicity.

I'm also too lazy to do this (mhmsmith0).

edit: actually if we're compiling stats from the last year might as well count the last year in the above too...

[Flubbernugget]

If you really want it you'll make it happen

[LicketyQuickety]

like 77% accuracy is just around random.

How is 77% accuracy about random?

[Ellibereth]

i lied and meant 75

assuming 10-3 is standard and you're town and randomly pick someone not you d1 and call them town you're right 9/12 so.

obviously when you're measuring this stuff you have to adjust for how many ppl alive and also ic's and confirmed roles/innocents/guilties since latter two aren't really reads anymore.

[LicketyQuickety]

i lied and meant 75

assuming 10-3 is standard and you're town and randomly pick someone not you d1 and call them town you're right 9/12 so.

obviously when you're measuring this stuff you have to adjust for how many ppl alive and also ic's and confirmed roles/innocents/guilties since latter two aren't really reads anymore.

It makes sense to me that in a 10:3 game D1 random chance for TRs is 66.67%. Why is it 75% and not 2/3 chance?

[RadiantCowbells]

Okay so you know for a fact that you are town

that leaves 9 possible town picks and 3 possible scum picks to a total of 12 picks

so 9/12 of people are town
3/12 of people are scum.

I think you're mixing up the math and doing 6 town 3 scum

[LicketyQuickety]

Okay so you know for a fact that you are town

that leaves 9 possible town picks and 3 possible scum picks to a total of 12 picks

so 9/12 of people are town
3/12 of people are scum.

I think you're mixing up the math and doing 6 town 3 scum

Yeah, and this demonstrate that I am not that smart either, because I did the math wrong and did 3/9 instead of 3/12. So I really have no idea why I have such a good record as Town (or Scum for that matter). Like, it would makes sense if I had some crazy high IQ and that is why my craziness works in games. But I don't actually have that high of an IQ. I think I top out at like 115 max, but it's prolly closer to between 105-110.

My point is that it's because I bring content to the game and do things that get stuff talked about. IDK why I have this ability, it's weird and I don't understand it. But there is too much evidence that says that I am a good Town player to ignore. Me bringing content to the game is really the only explanation.

[RadiantCowbells]

It's worth pointing out that your actual winrate on mafiascum specifically which is what's relevant given the amount of noise that other sites add is 50%.

Not to say that you're bad because having a 50% town winrate actually places you significantly above average in the current meta, but winrate is also queue dependent and being 0/5 in mini normals suggests that you're getting a lot of your wins in queues that are more favoured to town. I have a 58% town winrate on mafiascum and a 50% winrate in that queue, for example, and I largely avoid theme games where town winrates are the highest on the site.

Not to fail to give credit where credit is due because again being 50% puts you above a lot of people.

[RadiantCowbells]

Also I think that it's less content and amount of it you produce and more that you're resistant to town getting set in its ways and you stubbornly bring up whatever direction you're currently in. And yeah that gets you mislynched but I think it helps prevent town from getting complacent which is one of the big killers of towns imo. Being rarely NKed wouldn't hurt you either.