The Syndicate

S~V~S wrote:A sobering statistic~ we have mentioned Mr Piano Scanner 32 times so far. And I used the abbreviation for his name so as not to catch his quotes.

I don't know that he is that role, but he is trying to garner attention, IMO.

DharmaHelper wrote:

insertnamehere wrote:

DharmaHelper wrote:A turd with a top hat is still a turd.

I know, me giving reasons for things! I must be insane!

You're not insane. You know how I know?

I'm off my fucking rocker and I don't understand a word you say half the time.

DharmaHelper wrote:A turd with a top hat is still a turd.

DharmaHelper wrote:

Mongoose wrote:DH, if you know about a new train down the rail, let me know and I'll board it with you happily.

(And not in a "Let's trade murders" kind of way). Seriously though, I'll gladly talk about other stuff instead.

Lizzy's offering up her vote as I said is pretty darn pingy. Apart from her I'd have to comb through the pages of people talking about this whole as rey put it "Person of Interest" scenario and fuck that man I really am tired of hearing about it :P

Though I guess people's reactions etc. could be examined to determine suspects. Yes I realize the irony.

Linky: Because it's been talked in circles so many times.

INH: I have a program that can do X and Y and sniff out baddies.
Someone: Prove it.
INH: No
Someone Else: Show us some code.
INH: No
Someone Else: Use it on X
INH: No.
Someone Else: Tits or Gtfo
INH: Tits.

It's not productive. It's hardly discussion. :P

Epignosis wrote:

insertnamehere wrote:
It's only taking samples from the mafia threads, not the whole site.

How many Mafia threads are on ProgArchives?

S~V~S wrote:I agree with Goosey~ lots of people not playing this game that play a chatty game. LT would be a fine choice.

Epignosis wrote:

insertnamehere wrote:

Epignosis wrote:

insertnamehere wrote:

Spacedaisy wrote:INH, why not use Sockface instead of one of our current players to give us a sample. God knows he talks enough in every game he is in.

I contacted him, but he said that he could not, because it would influence the game too much.

Okay then. Do me.

How do I know you aren't bad? I can't take that chance.

The fact that you added Progarchives to your list is hilarious to me, in case you missed that.

It's a music site. I have 600+ album reviews over there. It isn't a Mafia site. What are you going to do with that data?

Epignosis wrote:

insertnamehere wrote:

Spacedaisy wrote:INH, why not use Sockface instead of one of our current players to give us a sample. God knows he talks enough in every game he is in.

I contacted him, but he said that he could not, because it would influence the game too much.

Okay then. Do me.

Spacedaisy wrote:INH, why not use Sockface instead of one of our current players to give us a sample. God knows he talks enough in every game he is in.

DharmaHelper wrote:

Bullzeye wrote:

DharmaHelper wrote:Anybody feel like, I dunno, moving on?

What else shall we discuss?

For a start, anything else.

Bullzeye wrote:

insertnamehere wrote:

Bullzeye wrote:

insertnamehere wrote:

Bullzeye wrote:
Prove it. Show us this difference, pick a player and blow our minds.

Why would I tell you guys the tells for a certain player in this game, that player could then avoid those words or phrases, and then his data will be inaccurate.

Why wouldn't you want to show that your system works?

I want to, but I want to do it without giving information to possible baddies in the thread, so they can break it.

So let me get this straight: You spent the best part of a year making a piece of software to analyse the posts of players (whose exact identities couldn't have been predicted) in one game so you'd catch their baddie tells and get a win for the civs, assuming that you yourself would win a game here and get a civ role. An expert in the field says what you're trying simply can't be done on this scale but that hasn't deterred you. Furthermore you can't show us how you made this system, any of the raw data, or its outputs in case baddies 'break it'.

Bullzeye wrote:

insertnamehere wrote:

Bullzeye wrote:
Prove it. Show us this difference, pick a player and blow our minds.

Why would I tell you guys the tells for a certain player in this game, that player could then avoid those words or phrases, and then his data will be inaccurate.

Why wouldn't you want to show that your system works?

Bullzeye wrote:

insertnamehere wrote:

Bullzeye wrote:

insertnamehere wrote:

Dom wrote: You "scanned" it?
What exactly does that mean?

I have been working on this for about a year. The Piano was the first website I scanned. I scanned it before it went down. I saved the comments of only certain people that were on The Syndicate, assuming they would be playing in this game. I only have copies of the comments of certain people in certain games of mafia.


Isn't the program already irrelevant if you told us about it at all? What if some people get nervous and screw up their civvie posting style? What if some people get nervous on the baddies, and they don't post normally either?
If this program really exists, have fun with it.

The reason I told everyone is because one person with a list in this game is completely useless. What was I supposed to do? Vote for the person at the top of the list without revealing why? Nobody else would vote for him, and he'd still be in the game. Once again, I am looking for subconscious tells.

So you made this software purely for the sake of this game, and started a year ago? Seems like a lot of time for something that will get on average 6 weeks use (assuming you're alive at the end). Also in my experience 'subconscious tells' tend to be physical things like someone twitching, touching something, itching body parts, eye movements, voice changes. None of this is conveyed through text. Do you have a source for the idea that subconscious actions can be picked up on over the internet? Normally people are very conscious of what they type. Typing is something you have to make a conscious choice to do. It can be influenced or inhibited if you're carrying out another language-based task at the same time but I don't think you're looking for people accidentally writing song lyrics or their shopping lists.

I agree in principle that there are behaviours that people show more when bad than good or vice versa, but these tend to be things that other players pick up on if they know them well.

This is supposed to be the big game on this site for this year, why not try it out here? There are such a thing as subconscious tells in writing, especially in games like this. There was a difference between the most used words by someone when they were good, and when they were bad, so I think I'm right on this one.

Prove it. Show us this difference, pick a player and blow our minds.

Bullzeye wrote:

insertnamehere wrote:

Dom wrote: You "scanned" it?
What exactly does that mean?

I have been working on this for about a year. The Piano was the first website I scanned. I scanned it before it went down. I saved the comments of only certain people that were on The Syndicate, assuming they would be playing in this game. I only have copies of the comments of certain people in certain games of mafia.


Isn't the program already irrelevant if you told us about it at all? What if some people get nervous and screw up their civvie posting style? What if some people get nervous on the baddies, and they don't post normally either?
If this program really exists, have fun with it.

The reason I told everyone is because one person with a list in this game is completely useless. What was I supposed to do? Vote for the person at the top of the list without revealing why? Nobody else would vote for him, and he'd still be in the game. Once again, I am looking for subconscious tells.

So you made this software purely for the sake of this game, and started a year ago? Seems like a lot of time for something that will get on average 6 weeks use (assuming you're alive at the end). Also in my experience 'subconscious tells' tend to be physical things like someone twitching, touching something, itching body parts, eye movements, voice changes. None of this is conveyed through text. Do you have a source for the idea that subconscious actions can be picked up on over the internet? Normally people are very conscious of what they type. Typing is something you have to make a conscious choice to do. It can be influenced or inhibited if you're carrying out another language-based task at the same time but I don't think you're looking for people accidentally writing song lyrics or their shopping lists.

I agree in principle that there are behaviours that people show more when bad than good or vice versa, but these tend to be things that other players pick up on if they know them well.

Kate wrote:

insertnamehere wrote:

Dom wrote: You "scanned" it?
What exactly does that mean?

I have been working on this for about a year. The Piano was the first website I scanned. I scanned it before it went down. I saved the comments of only certain people that were on The Syndicate, assuming they would be playing in this game. I only have copies of the comments of certain people in certain games of mafia.


Isn't the program already irrelevant if you told us about it at all? What if some people get nervous and screw up their civvie posting style? What if some people get nervous on the baddies, and they don't post normally either?
If this program really exists, have fun with it.

The reason I told everyone is because one person with a list in this game is completely useless. What was I supposed to do? Vote for the person at the top of the list without revealing why? Nobody else would vote for him, and he'd still be in the game. Once again, I am looking for subconscious tells.

I guess you missed my question, so I'll ask again. Would you kindly show us a screen cap of a scanned post from a game on the piano?

Dom wrote:Mongoose, do you think that role can be trusted?

Nevinera wrote:

reywaS wrote:Talk about entertaining. I thought the other game was funny, but this one is down right fucknutty. What is this, Person of Interest mafia? lol

My only ping so far was Nev's post giving advice to the baddies. That seemed weird, but Nev has posted many times and that's the only thing that struck me as weird.

My advice to the baddies was both good advice *and* good for the civs if the baddies follow it.

The thing is, Bob and the baddies are going to be at odds very quickly, and I'd much prefer that they realize that
each other are their first threats and leave off picking on civs for a while. Otherwise, the baddies will work against the civs
for the first 5-6 days and then Bob will have critical mass and beat them all using the inside information he will by then have acquired against them,
much like I used my knowledge via Epig against his team in bioshock.

This game, unless we are very lucky, will be about Bob.

Dom wrote: You "scanned" it?
What exactly does that mean?

I have been working on this for about a year. The Piano was the first website I scanned. I scanned it before it went down. I saved the comments of only certain people that were on The Syndicate, assuming they would be playing in this game. I only have copies of the comments of certain people in certain games of mafia.


Isn't the program already irrelevant if you told us about it at all? What if some people get nervous and screw up their civvie posting style? What if some people get nervous on the baddies, and they don't post normally either?
If this program really exists, have fun with it.

The reason I told everyone is because one person with a list in this game is completely useless. What was I supposed to do? Vote for the person at the top of the list without revealing why? Nobody else would vote for him, and he'd still be in the game. Once again, I am looking for subconscious tells.

A Person wrote:

insertnamehere wrote:

Nevinera wrote:Don't you see? It's a *proprietary trade secret*. If he revealed his code, well then *anybody* would be able to tell who's bad.

That would be tragic for the civ team of course. Clearly he's being civic-minded.

No, then the baddies could remake it, find out what their tells are, and avoid those words or phrases, making the Program irrelevant.

If I were a baddie I'd start having my teammates write each others posts, thereby making the program inaccurate.

Lizzy wrote:I for one am willing to let him dictate my vote on this first day since it mostly ends up in a mess. If you can produce a name that is the result of your programme at work, I'm voting for that person. If not, I'll go with the usual...

Nevinera wrote:Don't you see? It's a *proprietary trade secret*. If he revealed his code, well then *anybody* would be able to tell who's bad.

That would be tragic for the civ team of course. Clearly he's being civic-minded.

Nevinera wrote:

Bullzeye wrote: Wouldn't his libraries just be the sites that mafia is played on? I have no idea about this field so I'm guessing not, but it seems like the simple answer. I'd also like to know how he means to use The Piano as a source when it was shut down months ago and doesn't exist any more.

Software libraries. If he wrote his own machine learning implementation of any kind, I would *love* to see just that code of course,
but if he wrote a working document classifier as his 'first project', he's a genius and my company has a job waiting for him.

linki: No, I think you're a player that would like more control over the game. I do think baddie is somewhat more likely than civ for that behavior.

Once again, I'm new to this, this is my first project like this, so I don't know all of the lingo and the file types, but I think that it's not based on one of those specific file types that I can just search. The Program is specifically designed for word frequency in each game, and cross-references that list with other games. This simply is not something I know how to do, or probably can do.

And you claim to have written code that can parse content from forum sites and analyze it for common patterns? If you can produce your code, and it *remotely* works, I'll allow you to dictate my vote for the entire game.

Bullzeye wrote:

Nevinera wrote:I'm fairly confident that he doesn't even have such a tool and is just generating cover for future assertions and theories.
Even the task of scraping N different forums for data is a fairly complex one that one wouldn't just *do*.
That's why I asked for code and a list of libraries.

Challenge: if you have already scraped all that data into a db or files of some kind, it would be *trivial* for you to answer this question:
What are the top 20 words I've used in all threads on both boards I have played on under this name with length greater than 6?

Wouldn't his libraries just be the sites that mafia is played on? I have no idea about this field so I'm guessing not, but it seems like the simple answer. I'd also like to know how he means to use The Piano as a source when it was shut down months ago and doesn't exist any more.

Nevinera wrote:I'm fairly confident that he doesn't even have such a tool and is just generating cover for future assertions and theories.
Even the task of scraping N different forums for data is a fairly complex one that one wouldn't just *do*.
That's why I asked for code and a list of libraries.

Seriously? You think that I'm a baddie who invented a fake program to what, persecute civvies? First, that's the dumbest fucking idea ever, and second, if everyone I chose was innocent, and they were lynched, and then came up good, wouldn't I look really bad?

Challenge: if you have already scraped all that data into a db or files of some kind, it would be *trivial* for you to answer this question:
What are the top 20 words I've used in all threads on both boards I have played on under this name with length greater than 6?

Once again, I'm new to this, this is my first project like this, so I don't know all of the lingo and the file types, but I think that it's not based on one of those specific file types that I can just search. The Program is specifically designed for word frequency in each game, and cross-references that list with other games. This simply is not something I know how to do, or probably can do.

thellama73 wrote:

Nevinera wrote:

insertnamehere wrote: Well, since you have decided only to respond to that part, and not say how it actually will not work in your mind, but only automatically condemn me, is a bit suspect.

What part would you like me to respond to? I can hardly dump three classes worth of information and 8 years of experience in defense of my opinion on the topic.
Of course, if you would like to release a list of the libraries you're using and at least some of your source code, I'd be happy to look for bugs for you and try it on some of my company's sample data sets to see how well it does classifying customer demographics or twitter spam. You might have some chance of success at that latter one, if your code is correct.

I actually did say how it will 'not work in my mind', though. It will not work because you will have insufficient data.
I have no idea why you think you will have enough. You'd need hundreds of games worth of info from a single
person to make any kind of assessment about them with regards to any concrete variable with even a simple Bayesian filter - I doubt there is a single
person playing who has played more than 50 games of online mafia. And as for your assertion that *not* having a single variable means that it's less of a problem... that is completely backwards. having two variables instead of one would *square* the required sample size for any given confidence interval.

And *that* all assumes there even *is* a concrete variable in there representing 'writing style while bad'. Which there of course is not.

As someone who has studied statistics a bit, I would also add that your model assumes no change in style over time unrelated to baddie or civvieness. Even if you had hundreds of games worth of data, they would have to have taken place over many years. The assumption that players will not have refined and updated their style over time is unrealistic, and I can think of no way to adequately control for time in this case.

Nevinera wrote:

insertnamehere wrote: Well, since you have decided only to respond to that part, and not say how it actually will not work in your mind, but only automatically condemn me, is a bit suspect.

What part would you like me to respond to? I can hardly dump three classes worth of information and 8 years of experience in defense of my opinion on the topic.
Of course, if you would like to release a list of the libraries you're using and at least some of your source code, I'd be happy to look for bugs for you and try it on some of my company's sample data sets to see how well it does classifying customer demographics or twitter spam. You might have some chance of success at that latter one, if your code is correct.

No thank you, I don't fully trust you, why would I give you the resources to recreate what I did?

I actually did say how it will 'not work in my mind', though. It will not work because you will have insufficient data.
I have no idea why you think you will have enough. You'd need hundreds of games worth of info from a single
person to make any kind of assessment about them with regards to any concrete variable with even a simple Bayesian filter - I doubt there is a single
person playing who has played more than 50 games of online mafia. And as for your assertion that *not* having a single variable means that it's less of a problem... that is completely backwards. having two variables instead of one would *square* the required sample size for any given confidence interval.

Some players have over a thousand posts on the Syndicate alone playing mafia. Add in the other forums and you have thousands of posts from each player playing mafia. Once again, I don't intend for this to be set in stone, I just think that it's a better idea then randomizing, or some weak Day 1 suspicion.

And *that* all assumes there even *is* a concrete variable in there representing 'writing style while bad'. Which there of course is not.

Well, I think that all players subconsciously have tells while playing that differentiates their writing style based on their allegiance.

thellama73 wrote:I hesitate to say anything, for fear of being accused of baddie motives, but I don't like the idea of using a computer to play the game for us. It seems to defeat the purpose of the game to substitute a machine for our ability to read people.

I don't like it. I don't like it one bit.

Nevinera wrote:

insertnamehere wrote: Obviously, I'm not an expert at these things

Yes, my point was that I *am*, and I'm very confident that you will not get useful information from such a tool
(and I rather suspect that you have not written it either).

Bullzeye wrote:

insertnamehere wrote:

Bullzeye wrote:

insertnamehere wrote:So, let me preface this by saying that I am by no means, an expert at coding. This was the first really big program that I've decided to make. It could work, it could not, I'm going to try and see.

I've created a program that goes into every poster's game history, specifically games where they were bad, and sees what their top phrases or words are. It then contrasts those with the top words or phrases from a persons post history when they were good. It then eliminates the words that were on both lists, leaving us with that person's tells. After that, it goes into this thread, and sees how many times someone uses words or phrases identified as tells. The higher that number is, the more likely chance you are to be bad. This was a bitch to make, and it's something I've worked on for a long time, a bit of a pet project if you will.

Now, of course, because of the limited sample size, the program is more likely to be incorrect these first few days, but, as more and more people post, I know more and more about their role.

To summarize, I'm the NSA now, bitches.

The idea behind this sounds pretty interesting but I'm just not sure it'll work in practice. Can you think of anyone who specifically uses different words and phrases when bad or civ? It's more the attitudes and tones behind a post that change in my opinion.

Of course, attitudes and tones count for a lot, but everybody subconsciously has tells, and a lot of baddies like to fly under the radar, and not offend anybody. This is a way of finding those people.

Surely we can notice people flying under the radar based on what's happening in this thread? Also I'm sensing that this thing is something only you'll be able to use and any info the rest of us get from it will be the stuff you choose to share.

Nevinera wrote:

insertnamehere wrote:So, let me preface this by saying that I am by no means, an expert at coding. This was the first really big program that I've decided to make. It could work, it could not, I'm going to try and see.

I've created a program that goes into every poster's game history, specifically games where they were bad, and sees what their top phrases or words are. It then contrasts those with the top words or phrases from a persons post history when they were good. It then eliminates the words that were on both lists, leaving us with that person's tells. After that, it goes into this thread, and sees how many times someone uses words or phrases identified as tells. The higher that number is, the more likely chance you are to be bad. This was a bitch to make, and it's something I've worked on for a long time, a bit of a pet project if you will.

Now, of course, because of the limited sample size, the program is more likely to be incorrect these first few days, but, as more and more people post, I know more and more about their role.

To summarize, I'm the NSA now, bitches.

It's an interesting approach, but we have nowhere near enough data to do document classification for most of the players on the board.
I think you're much more likely to find patterns related to what other players are playing in the game.

I'm the lead developer at a company that specializes in pattern recognition and classification,
and while I haven't specifically written a data classifier, I'm familiar with the science.
Data sets for real classifiers need thousands of samples for 90% accuracy on things like 'author's gender', which heavily affect writing style.

I don't know whether you're just learning about machine learning and wanted to try something, or are attempting to grant yourself some scientific authority with which to influence future witch hunts or to cover up some other way of gaining info, but your tool is not going to be useful or effective, and is unlikely to beat random guessing in any statistically significant way.

Bullzeye wrote:

insertnamehere wrote:So, let me preface this by saying that I am by no means, an expert at coding. This was the first really big program that I've decided to make. It could work, it could not, I'm going to try and see.

I've created a program that goes into every poster's game history, specifically games where they were bad, and sees what their top phrases or words are. It then contrasts those with the top words or phrases from a persons post history when they were good. It then eliminates the words that were on both lists, leaving us with that person's tells. After that, it goes into this thread, and sees how many times someone uses words or phrases identified as tells. The higher that number is, the more likely chance you are to be bad. This was a bitch to make, and it's something I've worked on for a long time, a bit of a pet project if you will.

Now, of course, because of the limited sample size, the program is more likely to be incorrect these first few days, but, as more and more people post, I know more and more about their role.

To summarize, I'm the NSA now, bitches.

The idea behind this sounds pretty interesting but I'm just not sure it'll work in practice. Can you think of anyone who specifically uses different words and phrases when bad or civ? It's more the attitudes and tones behind a post that change in my opinion.

Epignosis wrote:

Mongoose wrote:

Epignosis wrote:

thellama73 wrote:

Epignosis wrote:
However, you are welcome to act wild and try to get the bad guys to vote for you, and then you either have information or the civilians in your wake have information. How did that work out for you?

Ah yes. You lost.

It was a moral victory. Plus, I got a maneuver named after me. You don't have any maneuvers.

That's why I have more wins than...

Okay, that's why I'm a mod here and you're...

Okay, that's why I know more about Mongoose and you...

You know what? You once sipped Red Stag out of a Days Inn paper cup. That makes you awful.

This post was so full of win.

I also LQTM at DH's most recent post.

I'm glad we are going to get down to business and vote for the most suspicious person, but I'm equally glad it looks like we are going to have a little fun doing it along the way.

I already have someone in mind who will be getting my vote but I'm waiting to see what he does.

S~V~S wrote:MP did something not totally dissimilar to that once~ he blew up ALOT of players on one night with bombs.

reywaS wrote:ha

Vompatti wrote:

Mongoose wrote:Vomps, tell us about your vote, son.

I don't want to talk about it.

S~V~S wrote:I went with twin peaks. I had NO CLUE WTF happened in that game, but i played it well. it was one of my better ones.