Mafiascum.net

In post 30, mutantdevle wrote:Tbh, I'd probably win more if I just didn't have reads. If you were to compare random to lynching to only lynching my scum reads then random lynching would be more successful :/

Don't sweat it! You're still improving.

One of your greatest strengths that I've seen in your towngames is your ability to utterly spew town in half your posts, removing yourself from the lynchpool. I probably should've caught that you were scum in Open 714 (Tit for Tat) sooner just by the fact that I wasn't hard-townreading you.

Early scumreads are basically always a shot in the dark unless you know the players around you really well (or the scumteam is bad). What I would work on is figuring out why your townreads were wrong. What is scum capable of faking on this site? What kind of logic/progressions are actually deserving of townreads?

I have a few answers, but as always, publishing tells all over the place is a great way to invalidate those tells. You pick those kinds of things up over time, whether by yourself, or by listening to the logic of other players.

So in the meantime, recognition that your early reads are often flawed is a great foundation for
- Avoiding overconfidence
- Working on what your strengths CAN be in early games
Try sticking to reading the players you know, or trying to rule out scumteams so you know when to sheep other players (see the Know Your Weaknesses section).

In post 41, GuiltyLion wrote:yo MD this article is great and overlaps a lot with the way I try to play the game, even put some ideas into words that were kicking around somewhere in my head but that I hadn't considered so explicitly before. Thanks for writing it

Thanks!

In post 42, GuiltyLion wrote:also make sure to Bayesian analyze individual posts

like if you see a post that is absolutely super unfathomably unlikely to come from scum - such that Probability(Player made post [x] given that Player is scum) is ~0, then it doesn't really matter how much decent evidence you had before for them being scum, they're probably town. I have the highest success for accurate townreads when I find good towntell posts like this.

Same thing goes for scumreads if you find posts that are almost impossible to come from town. Unfortunately, those posts are more rare both because people are trying their hardest to avoid making them and because generally in your experience (and in each game), the volume of townposts >> volume of scumposts

hell yeah my dude

i might make a post that's full on solely about bayesian reasoning in mafia, i don't see that much in the MD

In post 50, Mathdino wrote:I have a few answers, but as always, publishing tells all over the place is a great way to invalidate those tells. You pick those kinds of things up over time, whether by yourself, or by listening to the logic of other players.

Certain things are easy to fake, certain things are hard to fake.

The manner in which people try to solve the game, the depth and quality of reasoning, the emotional investment, the attitude of "aggressive curiosity" (as it was put to me by a veteran player on another forum), etc etc... most of these things are difficult to mimic.

Angry yelling is easy to mimic. Pushing on wolves (i.e. distancing/bussing) is easy to mimic (though sometimes dumb to actually do as a wolf). etc etc

In post 48, mhsmith0 wrote:

In post 46, Mulch wrote:

In post 45, callforjudgement wrote:

In post 30, mutantdevle wrote:80% of my null reads are town.

That seems about right. Most setups have around 80% of

players

as town. So if you have no information about a player (or think it all cancels out), they'll be about 80% likely to be town.

I can’t tell if this is right or wrong

Someone smart help me

Who can know if the path we choose is right or wrong? I will fear no reproach. For man is a creature of sorrow and regret, and ever will be.

I actually think it’s mathmatically incorrect but I can’t articulate why

Well, it largely depends on how you define "about"

10/13 is 77%
7/9 is 78%
etc

OTOH if you're tonw, then you're really concerned with the OTHER players, and then it turns out to be exactly 75% for all standard numbers formats (this is because standard numbers has T = 3*M + 1, and if you take yourself out, it's just T (excluding you) = 3 * M)

In post 46, Mulch wrote:

In post 45, callforjudgement wrote:

In post 30, mutantdevle wrote:80% of my null reads are town.

That seems about right. Most setups have around 80% of

players

as town. So if you have no information about a player (or think it all cancels out), they'll be about 80% likely to be town.

I can’t tell if this is right or wrong

Someone smart help me

Could Something Smart help you?

iirc around 22%,are scum. but he's wrong, because it's around 66% excluding yourself

take a 10:3 mini
9 other slots are town
3 other slots are scum
thus it's 75% to 25%

I think it’s because in practice your null reads town chance should be affect by number and strength of your town and scum reads

This was a good read, thanks Dino et al. I have very little to contribute, but it's good to get ideas on how to improve.

You dont need good reads to win games and I disagree with your first section where you say its the first skill one should learn.

In reality good reads are the hardest skill and you can win games without them.

Like there are a fuck ton of skills required to win a game of mafia and good reads alone wont win you a game.

In post 60, beeboy wrote:Like there are a fuck ton of skills required to win a game of mafia and good reads alone wont win you a game.

Terrible reads will lose you games tho, especially if they’re high confidence ones.

In post 61, mhsmith0 wrote:

In post 60, beeboy wrote:Like there are a fuck ton of skills required to win a game of mafia and good reads alone wont win you a game.

Terrible reads will lose you games tho, especially if they’re high confidence ones.

If your hard pushing horrible reads I would argue you lack the other skills as well

Maria, Alisae, Mastina, NSG, and RC are all examples of (different levels of) good by current site standard players and I've seen hard pushing horrible reads in multiple games. Everyone has off games.

In post 59, beeboy wrote:You dont need good reads to win games and I disagree with your first section where you say its the first skill one should learn.

In reality good reads are the hardest skill and you can win games without them.

My intent wasn't to argue that having good reads is the first skill to learn. I think in writing this article before "How to Read People" (something I probably won't write) I'm arguing that Self-Appraisal is a skill that should come first

My intent was to argue that scumreading scum and townreading town is the first step toward lynching scum and not lynching town ingame.

Personally, I developed towntelling and charisma long before I developed good reads. Not getting lynched is great and all, but one more Innocent Child with so-so reads just doesn't swing games in the current meta, in my experience.

Knowing how to listen to people and not impulsively scum reading them, and or lynching them, for having different reads than you can go a long way to improving your game play.

I mean I'd take someone consistently okay vs. someone who swings between bad and great anyday.

Good reads are a way to "win" the game, I think if most people want to improve they need to learn how to "not lose" games first.

In post 63, Eddie Cane wrote:Maria, Alisae, Mastina, NSG, and RC are all examples of (different levels of) good by current site standard players and I've seen hard pushing horrible reads in multiple games. Everyone has off games.

I actually push players expecting to be wrong, for what it's worth--I just think that the information gained from a wrong push is better than whatever is gained from not making the wrong push. (This doesn't mean I don't believe the push. I do! I'd never push something I don't believe. It's just that my self-expectation is that my belief will be wrong. I just don't think the game works if you always assume you're wrong. I expect to be wrong, but assume I'm right; the two aren't synonymous.) Thus my Push Hard philosophy.

Also I'd argue that I'm not good. I'm average, for exactly that reason; I do in fact have off games and even when I have games where I nail like three scum I have faults keeping me from having done perfectly.

i've never self reviewed. I stick my completed games on my wiki so lets do that

1976 - I knew something was off with Derpy Hooves, he turned out to be the serial killer. I had a boner for BuJaber as he reminded me for someone who is lynchbait on another forum that i play on - that was wrong. I also wanted to PL mulch and he was scum. This was a weird game because I was trying to enact a PL towards least damage to town, so not a great read as such, even though I was hitting anti town players.

1842 - I remember making a very contrived case against Not_maf and UCV in this game but we eventually caught the scum quite early in the game. I can't remember how it came about, I dont think I can take a massive amount of credit in it, but looking at my ISO somehow I convinced Leeoon to hammer eth0s for a win... I think more luck than judgement as i was still learning site meta.

1982- I think this was the game where I picked up on a mafia frame kill very quickly and chumba denied it would even happen. I was still learning chumba meta at this point as I was still frustrated with his derpy hooves alt play in the first game i did here. I think i also voted mulch for pre-game bias from 1976 too. I could not read Tchill at all and alienated him so badly that he wouldnt work with me so I take massive credit for causing a town loss there, I was so focussed on using my JK role to solve the game perfectly but tchill wouldnt work with me because of the way i went about it - it took me a real life sleep on it to realise he was town too. I got there eventually but it was too late. Again a hit and miss game for me.

1984 - we lost this one too. I remember town reading sauce despite being the most obtuse player i've encountered but this caused me to misread thor. i played a significant part in misreading Flavour Leaf and he replaced out but i correctly town read his replace in. There was a cursed scum slot that had mega inactivity which I'd like to say helped their victory (sorry if any of those players are reading) Acryon did really well though, i defo didnt scum read him so not a great game for me.

786 - a weird game where each player had a PT with each other, I employed a gambit which got my reads down to a correct town block and I would only have lynched in my scum pool for a win as there was a max of 2 lynches in the game. as it happened eth0s shot from the hip and we insta-won. On reflection I realised that Mathdino offered up another gambit where he said let's lay some votes down and see if someone lynches. The only way he could have secured a non-lynch is if he knew scum wouldnt hammer, the only way to know this was he was scum - hindsight is wonderful - but - i would have lynched him day 2 so either way, I'm happy with my read of the game here.

1851, I remember picking up the Gem read pretty early on but couldnt get an early lynch on that slot. I also had a case of paranoia over NSG and couldnt quite fully town read her but got there in the end. I think my reads were just about ok but could have easily gone the other way here. I think my biggest reflection on this game is that i should have pushed Gem hard earlier to make the remaining scum slots life much harder.

1856: i replaced in and scum read thor almost straight away, I remember having paranoia over chumba because he always says he cant be read and he was playing the same game he always plays, so i was dancing around that a bit. I didnt get DDS but we won via PR help quite easily.




Overall I think I can pick out from my games memories of distinct reads and they were usually right. The self-lesson here is when I think I'm right, just run with it and sort the other slots.

I seem to have a general knack of surviving quite a long time which im not sure is always positive and i feel like if I'm certain on my scum reads i can get rid of those players.
I'm not a great scum hunter when i am not sure on the slot so i need to work on that

In post 66, Ellibereth wrote:I mean I'd take someone consistently okay vs. someone who swings between bad and great anyday.

Good reads are a way to "win" the game, I think if most people want to improve they need to learn how to "not lose" games first.

Ya Eli I still think your good even if your only just ok every game

In post 66, Ellibereth wrote:I mean I'd take someone consistently okay vs. someone who swings between bad and great anyday.

Good reads are a way to "win" the game, I think if most people want to improve they need to learn how to "not lose" games first.

Definitely, no question. For example, someone who's perfect 40% of the time but awful 60% of the time when they're town is not worth betting the game over. Because there's only a 75% chance that they're even town to begin with -- that comes out to a 30% probability that what they say is right, and I know from experience I can, in general, do better than that.

That said, the point is that "great" isn't necessarily in the quality of someone's reads, but in the player's ability to evaluate the quality of their own reads, and use that as a way to both not fuck up in game, and improve as a player out of the game. If you know what you don't know, you're halfway there.

I think I disagree with the pessimistic take that townies nowadays should just take a step back and let the known good players just determine the outcome of the game. For one, it's no fun and isn't really my view of what a good game of mafia looks like, and for two, it seems kind of defeatest for the player's future skill level. Players with bad reads can absolutely evolve to be town leaders with good reads. I think I improved as a player the absolute most in 2014 after pushing mislynches and realising where my reads went wrong.

But you have to have those realisations!

It’s interesting that this thread has largely focussed on scum reading and ignores the skill of identifiying potential PRs: A) whilst playing scum and B) whilst tying to protect them as town

The easiest and most reliable way to identify power roles is if both of these are true: a) the player claims they're a power role, b) the setup doesn't make sense balance-wise unless they're a power role. As a bonus, this often lets you figure out their alignment too.

This sounds trivial, but I'm still surprised how often people get just so sure that someone else is scum that they continue to consider them scum even when setup balance evidence points in the opposite direction.

This is a good thread. I generally approve of self reflection. I think it would be interesting to review your own recent games, with someone who is equal to or beyond your skill level. You go over your games together, and comb over your strengths and weaknesses. Go over bad habits you should break, and brainstorm how you can improve your weak areas and boost your strong areas together. It could even be a mutual review if both players have close skill to each other. Basically, like a hyperbolic time chamber, or deep training as I would use for super smash bros or any fighting game. (Three people take turns 1v1'ing each other, the person out of the current rotation takes notes of the play they have noted, and after the set, they reflect on the matches and then rotate)

In post 71, profii wrote:It’s interesting that this thread has largely focussed on scum reading and ignores the skill of identifiying potential PRs: A) whilst playing scum and B) whilst tying to protect them as town

That's actually one of my best skills!

I have a draft saved for an article about doing just that: PR/VT reading.

In post 73, Ranmaru wrote:This is a good thread. I generally approve of self reflection. I think it would be interesting to review your own recent games, with someone who is equal to or beyond your skill level. You go over your games together, and comb over your strengths and weaknesses. Go over bad habits you should break, and brainstorm how you can improve your weak areas and boost your strong areas together. It could even be a mutual review if both players have close skill to each other. Basically, like a hyperbolic time chamber, or deep training as I would use for super smash bros or any fighting game. (Three people take turns 1v1'ing each other, the person out of the current rotation takes notes of the play they have noted, and after the set, they reflect on the matches and then rotate)

This is a really great idea!
I don't really have anything to add to this point other than that I definitely approve
Go teamwork