Mini 515 - The Pine Barrens - Game Over!

[Adel]

I'm here to replace Zakarum.

I've played with BooKitty and ThAdmiral before, and Elias_the_thief a few times before.

[Adel]

a few questions for all of you, to help me with my game read:

1. who are the active players who make content-rich posts?
2. who are the lurking players, who post just enough to avoid attention?
3. which players post many words, but little information?

The more people who answer these three questions, the quicker I'll be able to contribute.

Thanks, I'm looking forward to an interesting game.

[Adel]

Thanks for the welcome, could you answer these questions, please?

1. who are the active players who make content-rich posts?
2. who are the lurking players, who post just enough to avoid attention?
3. which players post many words, but little information?

It really will help me.

Thanks.

[Adel]

I've been looking into voting patterns lately. Currently I throw out all posts from the random stage as noise, and start my analysis from there.

I consider the end of the random stage to be Boggzie's vote for ryan at post 63

here is the game action's I recorded. please check my work: I am especially prone to missing unbolded Fos's.

63 - boggzie votes ryan -4
64 - ryan votes boggzie
68 - mcpaltp votes ryan -3
70 - hasdgfas votes boggzie
72 - opie votes ryan -2
82 - hermit unvotes ryan -3
82 - hermit fos boggzie
85 - bookitty votes ryan -2
86 - elias unvotes ryan -3
87 - spider fos ryan --- _{note that spider did not vote ryan before ryan was moddkilled. spider had every chance to, but didn't. this is an example of what I consider to be a typical townie vote pattern. fos's are not followed by votes. scum, on the other hand, tend to fos a target townie, and then vote later hoping the prior fos will give some rhetorical cover for their vote. scum almost never fos a buddy before voting him. they will, however, point a fos at a buddy (especially a lurking buddy) but will not vote for that buddy until a late bus is needed.}
98 - elias fos bookitty
99 - elias fos mcpaltp
106 - elias votes bookitty
107 - hermit votes bookitty *** _{here hermit, after making an earlier fos at boggzie, votes for bookitty.}
108 - bookitty unvotes ryan -4
109 - Zakarum votes bookitty -4
[quote=votecount 4, post 110]
ryan - 3 (Boggzie, mcpaltp, opie)
Bookitty - 3 (Elias_the_thief, TheHermit, Zakarum)
Boggzie - 2 (ryan, hasdgfas) [/quote]
ryan modkilled at post 146
Spider nk'd
154 - mcpaltp votes TheAdmiral -5
155 - hermit votes mcpaltp -5
156 - zak fos mcpaltp
167 - elias vote bookitty
167 - elias fos boogzie
181 - elias fos hasdgfas & opie ^^^ does elias seem scummy for sending out

that

many fos's?
183 - zak fos hasdgfas & opie ^^^_{zak is an idiot for following elias like that. regardless of role or alignment, following another player like that is an idiotic move. shame on zak. this post alone earns him the title of "village idot" in my mind.}
190 - bookitty fos zak & TA
190 - bookitty vote mcpaltp
218 - opie fos zak
224 - neko vote TA -5 *** _{scum votes buddy for distancing?}
230 bookitty vote TA -4
240 - neko unvotes TA - 5 **** _{scum abandons wagon of buddy before it gets too large?}
243 - neko votes boogzie
245 - hasdgfas votes zak
262 - mcpaltp votes boogzie
266 - neko unvotes boogzie
266 - neko fos mcpaltp
284 - neko vote mcpaltp - _{does regular voting following a fos still count as a scumtell?}
304 - opie votes zak --- _{another vote following an fos. this time opie even points to the earlier fos for cover}
308 - hermit votes neko *** _{possible scum distancing vote.}

[Adel]

[Adel]

For Mafia I use Fireworks, Mathematica, and a text editor. I created those graphics myself.

I wouldn't worry too much about those notes. They are just surface-level observations I made while collecting data. The deeper analysis takes much longer. The post where I present my conclusions will include a vote.

[Adel]

vote: opie

fos: Bookitty

Based on my model of scum voting actions for their respective activity level.

I will not share the algorithm I use, or the data points I rely upon. I have about 70% confidence in opie being scum, and I have about 67% confidence in a Bookitty lynch being accurate. I do not have any significant evidence connecting them as buddies.

[Adel]

active games, so no.

[Adel]

I find the graphs to be a useful reference. Sometimes other townies find them useful as well, usually those townies who have visually & spatially biased intelligence. They are not meant to point anything out. They are part of the data set I look at when I start my analysis.

I just noticed that I didn't answer Bookitty fully with my last post: my success rate with sample mini-normals games where I use the first 3/4's of day one for data, is almost 80%. This game, with the modkilling, and the unusually uniform contribution level, may not fit my model as well as I would like. So I have less confidence than I usually do. There are no completed games where I used my algorithm- I only recently finished it. Mini 458 (see my wiki page) is a game I played in as scum where I did reference models and used graphs. ThAdmiral was a lurking & townie player in that game.

One of my assumptions is that there are 3 members of the mafia, and only a 15% chance of a SK. So the baseline % chance of a specific player being scum is ~39%. I treated all of day 1 and part of day 2 as all being the same day, and I did not take ryan and Spider's revealed alignments in to account in any way. Thankfully, my model did not predict either of them as being scum.

[Adel]

:
Is Post 335 the extent of your deeper analysis and conclusions?

no, it is the

result

of my deeper analysis

Are you

only

voting and (FOSing) based on scum voting action for my respective activity level?

no, your activity level only determined which model I used to evaluate you

Could you explain the logic behind this model.

The votes on each player act as a sort of a market. Once a player meets a majority of that market's share, he is lynched. The technical analysis used to predict changes in some commodity markets is very similar, mathematically, to a mafia game. This is a simple example of what influenced my design.
The nivel aspect is that the goal is not to predict how large a wagon will get, but which actors use the mafia equivalent of

insider trading

to determine which wagon they will support.

What assumption did you make when you developed this model?

many, many assumptions, which I will not be sharing. The central assumption is that the more assumptions I incorporate, the more robust my system will be, which is assuming that more of my assumptions are accurate for a given game than not.

Could you plug every player into your model and post their results?

I did plug every player in, and I will not share the results

Why can't you share the algorithm you use or the data points you rely on?

I want my system to remain accurate in future games. If I released that information the meta would shift, and my ability to accurately identify scum would rapidly decrease

What's your percentage of confidence based on?

if a townie did something in 12% of my sample games, and scum did the same thing in 56% of my sample games, I derive a value from that, and plug it into an equation.

Is your vote for me in part based on my FOS and then voting for Zakarum?

no

But you go on later to say that scum will point a FOS at a buddy, but

not vote

for that buddy until a late bus is needed. You are implying then that typical scum behavior is: FOS (for buddys) not followed by votes (unless bus). If you discount our hindsight (i.e. before the end of Day One), couldn't one have used this logic against Spider Jerusalem? Couldn't one argue based on this logic that Spider pointed a FOS ryan (his buddy) and not voted for him? And was unable to bus his buddy because ryan modkilled himself first?

This is the kind of paragraph that led me to try to develop an accurate model based upon objective observations. It is all bull in my eyes. Not that that makes opie any more scummy to me, pretty much every player is so full of bull I have a lot of trouble telling scum from townie.
Based upon a large number of games, I've noticed a couple of voting patterns that

generally

hold true. That kind of analysis is still too intuitive for me to build an algorithm around it. I would like to.

Bookitty, I would not be expecting a vote from Adel. Here's why:
(a) He pointed a FOS at you in Post 335;
(b) Adel claims to be town;
(c) Townie's don't follow a FOS with a vote for the same person.
Therefore, I wouldn't worry about a vote from Adel. For if Adel

were

to vote for you then he would be acting contrary to typical townie behavior and thus likely scum.

is opie warning his scum buddy not to overreact to my post?

[Adel]

I think the graph is really fun to look at, and the summary of events was useful to some extent. But I really don't think it's any substitute for actual analysis of what is said, and who said it, and why. If it were, then computers could play the game, and people wouldn't be necessary.

I thought Zakarum was scummy before Adel arrived, and her latest postings seem very distractive and not very helpful to town, even though they're visually stunning. Basing your vote on wordcount? It seems really counter to the spirit of Mafia, and hopelessly random. If she could have pointed to instances where her theory was proven correct, I might have felt differently, but at this point, it just looks like a lot of razzle-dazzle and distraction, and not really helpful to town, especially considering the (in my opinion, at least) scattershot results she seems to be deriving.

Based on my own impressions being nearly diametrically opposite to Adel's results, and on Zakarum's past play, I will:

unvote; vote Adel

I consider this to be the overreaction opie was warning against.

[Adel]

another scum warning.

FoS: mcpaltp

for trying to warn a scum buddy against over reacting. I have linkage between mcpaltp and opie

or

Bookitty. I dubt all three are in the same scumgroup though- my model is terrible at predicting all 3 members. If Bookitty or opie are lynched and revealed as scum today, mcpaltp is my first choice for lynch tomorrow. He is

not

a good choice for today's lynch though.

[Adel]

"leading the town" is one of those stupid observations that is plaguing mafiascum right now. Like political rhetoric it is a negative "talking point" that isn't based on anything of substance, yet manages to create a negative impression in other players. No-one likes being told what to thing, and you post was designed to create that impression.

I was sharing my conclusions- stating my opinion.

The scatter point graph does help to reveal that your typical posting style is little quips like the above one. I haven't figured out how to accurately quantify it, but it strikes me as scummy.

Do you like applying little pokes and prods without drawing much attention?

[Adel]

While Adel's methods may be suspect, I've seen some very interesting behavior since the posting of his graphs and his subsequent discussion and vote. There seems to be a lot of...overreacting to his methods and I wonder about that. It seems like they are only worrying about him instead of the group as a whole. I don't know why, since, at least imo, the graphs are nice, but don't help me in any way since I don't know the logic behind it. I'd rather have some hard evidence rather than graphs that I don't know the meaning of.

The voting diagram is easy to use as a kind of a map. If you want to lookup the post where opie voted for zak, for example, it gives you the post number.

Another way to use it is to print it out. Assuming that you are town-aligned, you know that three names on the map are innocent. Take an orange highlighter and trace over the FOS's and votes pointing at those innocent names.

If a player is vote hopping, or following the lead of another player, the map makes it easier to see.

The scatter point graph is more simple. I usually group players into three categories: active, passive, and lurking. The scatter point graph makes it easier to determine which category each player falls into. This game is a little unusual because Bookitty is the only player that is obviously active. Generally, there is a clear gap between the active and the passive players.
So I had to run my system a few additional times, and determine a way to weigh the difference in scores. Not a real problem.

Lurkers also show up very clearly on the scatter point graph.

[Adel]

it is rather hard to lead the town when you are pulling the largest wagon, as I am.

[Adel]

thanks for the spam.
That is a great illustration of how your data

can

will get corrupted if the scum know what data you are collecting.

I tend to either get lynched the day I replace in, or I get Nk'd the following night. I am very used to it. My major problem has been figuring out how I can get the other townies to remember my posts after I am dead.

[Adel]

It seems very pretty... but it's rather hard to understand!

You see she didn't like to confess even to herself, that she couldn't make it out at all.

Somehow it seems to fill my head with ideas--only I don't exactly know what they are! However, somebody killed something: that's clear, at any rate...

I am fond of children, except boys

otherwise...





D'accord?

[Adel]

fine then, a reference to

Jabberwocky

along with a hint that it includes a breadcrumbed role claim, in addition to the

Pine Barrens

flavor, leads me to conclude that mcpaltp's post @ 361 shows that he is a werewolf.

So, despite my earlier post saying that he was not a good lynch choice for today,

unvote:opie, vote:mcpaltp

neko, does that leave you feeling any happier?

[Adel]

A couple of games Rishi may have consulted if he were designing a game with two scum groups.

monks and masons
Too much scum

Worth a read.

[Adel]

opie has 57 game posts on mafiascum.net

Now he is telling me how to play mafia? Seriously, STFU with that "spirit of Mafia" nonsense. Mafia is a game that was developed by Math geeks with an interest in game theory, like me. We tend to ignore 2000 word posts filled with subjective nonsense and logic so fuzzy and undefined that it is impossible to engage.

Since I've replaced into this game I'll hazard that I've put more time and effort into it than the rest of you combined.

Because you do not understand my posts, you assume that they are not pro-town. Spend some more time looking at them, and at least

try

to understand what I am saying.

another link for you to follow: This time it is a game modded by MeMe that attracted a number of very experienced players
Note that I was lynched the same day I replaced in, and I was a townie. Try to follow my posts- I did generate a game winning tactic (rooted in math) that the town decided to lynch me for. Pity if that were to happen again.

[Adel]

Ugh. This must have been posted while I was composing my post.

I'm slightly baffled my this. I see mcpaltp suggesting a breadcrumb, but where are you coming up with werewolf? And if he was werewolf, why would he claim now?

it would be a cute and smug thing for him to do, which fits my very rough and amateurish psychological profile for him.

And now that you are following a FOS with a vote, should I consider this a scumtell on you?

thinking along those lines leads to recursive logic errors. Good luck being successful at mafia.

[Adel]

if you have an interest in breadcrumbing, this thread contains a few interesting ideas.

[Adel]

the jabberwocky is a big monster in the woods that you have to kill with a magic sword, found in a poem that is am example of literary nonsense. mcpaltp is a fountian of nonsense claiming a role with a poem about the jabberwocky. the flavor of our game would allow for the presence of werewolves.

i think mcpaltp spends a lot of typing to amuse himself at his own cleverness, and hinting at his scum role in such a obtuse way would give him a case of the giggles, i'm sure.

[Adel]

A point of clarification: I have never run my system in a game that has completed as of this date.

[Adel]

a mini that just ended tonight because of the unusual mechanic, I used a simplified (i.e. quick and dirty) version of my system. The first lynch worked out fine, but once the game came down to pdcakes and JDodge my system told me that pdcakes was the guilty one. I almost voted him too.. but I don't have blind faith in my system so the town won the game.

As a few of you have suggested, I didn't mention my system in that game, and I didn't use any graphics. I took the results my system suggested and then tried to develop a written case against that suspect. I though I would try being a little more honest in this game- and actually explain what is driving me to the conclusions I reach.

[Adel]

I'm in a game full of SA players. oh. lynch me now.

vanilla townie

here, not full of noise, but at least it is safer than continuing with the day and possibly outing a powerrole. I've said pretty much all that I want to in this thread for this audience anyhow.

unvote: vote:adel

[Adel]

a note about scumtells: they are just like tells in poker. If I know you rub you eyebrow when you try to decide if you want to bluff or not, I will beat you in poker on a very regular basis. If you figure out that I am looking for if you rub your eyebrow or not, you will have the advantage.

I keep the tells that I look for secret so that scum will not adapt.

I agree with the general sentiment of not trusting Adel and his magical mystery algorithm. Especially when he asks us to just trust it, then admits that it screwed up in the recently completed game. A-durr.

I won that game. A-durr. I know when to use it and when not to use it.

Everyone and their mother has a "system" for winning it big in Vegas. They all fail, spectacularly. Do you know why that is? Mostly because it's essentially impossible to predict randomness, but also because math ignores the human element, and you simply can't do that in a highly personal game like mafia.

counting cards when they only used two decks for blackjack used to be a system. automated investing strategies are common on Wall Street, especially with Hedge funds. to clean up NYC, an algorithm was used to determine police beats, with great success. "the human element" is a myth, for any system there will be patterns that can be modeled and predicted with some degree of accuracy. even a totally random system will produce results that are not totally random.

[Adel]

look, i think it is pretty obvious who has been trying to get me lynched. there are two scum on my wagon: opie, mcpaltp, and/or Bookitty

I am confident that I identified the scum.

A mislynch of me leaves 8 alive tomorrow. 5 town & 3 scum

A correct lynch of opie tomorrow 5 town & 2 scum

followed by a nk: 4 town & 2 scum

followed by a lynch of either Bookitty or mcpaltp: either 3 town & 2 scum (scum win) or 4 town and 1 scum (probable town win)

a this point, as I see it, my lynch is pretty much inevitable. the only question is how much noise and confusion will the scum be able to sow before my death scene? I hope to minimize that.

good luck town, but I don't have a very high expectation of us winning. at all.

[Adel]

lynch opie tomorrow, and if that is a miss, vote boggzie to get out of the lynch-or-lose, then consider mcpaltp against Bookitty for the town win.

[Adel]

also, when you do your own analysis, you now know four names of innocents: me, ryan, and yourself. keep that in mind as you use an updated version of the voting diagram.

voting diagrams, once you learn how to read them easily, can be a great study aid.

[Adel]

who knows how long this twilight will last, but I am online, checking this thread continually, and i'm willing to answer questions...

[Adel]

Good job scum -- what an overwhelming victory!

[Adel]

well, my graphs weren't much good for this scum group.

[Adel]

I agree -- he seemed to be more astute than the average player.

[Adel]

by the way, this is the second game (out of 17) I've played in where ryan got modkilled for doing something clearly against the rules. He is on my blacklist -- I will not join any game where he is a living player.

[Adel]

And I'm very sorry the town had to die in order to solve my self-esteem issues and allow me to feel accepted. But at least I lived happily ever after until Fat Tony shot me in a territory dispute one week after my induction into the mob.

-- that left me expecting some great flavor from some future game modded by you.

[Adel]

I think we were also really distracted and ultimately harmed by Adel's methodology. Not only were his results completely incorrect, it also really shifted focus off of the business at hand.

Granted I was totally wrong with my analysis, but I was pretty shocked at how much negative attention I pulled.