Ozy's Scumhunting Spreadsheets - 2021 edition
Register

User Tag List

Results 1 to 9 of 9
  1. ISO #1

    Ozy's Scumhunting Spreadsheets - 2021 edition

    In my tables/spreadsheets that I used to scumhunt in the Silent Night game of 2019 December, it was mostly focused on Activity and Effort. I wanted to update my spreadsheets to something more diverse even back then, but I couldn't.

    I attempted to do it now and I think I succeeded in improving them.
    Wanted to share these and ask for your opinions on them.

    I have been wanting to test them for weeks now by going through past games and checking if scum are being found with it. I could never get myself to do it.


    Spoiler : Attributes :

    Spoiler : Town Attributes :

    Encouraging Content - Any “random” comment that can make people comment on it will mark the slot with a YES in this attribute. I reckon that Mafia don’t have an incentive to push the game forward.

    Making New Points - Any new/original (non-mechanical) idea or point will mark the slot with a YES. I’d imagine every slot to get marked with a YES here, especially because I cherry picked games where all slots are decently active.

    Providing Conclusions - If the slot has provided any conclusions based on actual arguments, they’ll be marked with a YES.

    Lack of Self-preservation - If the lost doesn’t “feel” to be careful, they'll be marked with a YES.

    Confirmation Bias - Alignment Reads wise, it’s impossible for Mafia to have a Confirmation Bias. Therefore if the slot sees any random stuff as a confirmation of one of their ides, they will be marked with a YES.

    Alignment Paranoia - Same as above, Mafia can’t have Alignment Paranoia. Therefore any “excessive” distrust will mark the slot with a YES.

    Paranoid OMGUS - Technically the same as above except it’s OMGUS, therefore only OMGUS counts here. Had to be put in the list in order to differentiate from the “I’M INNOCENT, YOU SUCK!” OMGUS that’s in the Scum Attributes.

    Cooperating - Cooperation and Buddying need to be told apart. Therefore, just like with OMGUS, one is in Town Attributes while the other is in Scum Attributes, while having neither is neither scummy nor towny.
    Spoiler : Mafia Attributes :

    Apathy / Coasting - If the slot doesn't seem to care about the game or makes low effort / has low activity, they will be marked with a YES.

    Sheeping consensus - A slot might make new points and conclusions and yet still Sheep the threads consensus and follow the least controversial ideas. I expect an unfairly large amount of slots will be marked with a YES here.

    IIoA (information instead of analysis) - Any big post that has a lot of factual info in it, but no analysis or conclusion from it will award the slot with a YES in this attribute.

    Gaslighting or Gish gallop - Excessive amounts of AtE, arguments or ideas will mark the slot with a YES. Arguably subjective and for some players that’s just how they play, not to mention that I’ll check Light Games which makes it a low possibility that any slot will be marked with a YES.

    Disassociated - If the slot feels like they are disconnected / distanced / impersonal, they’ll be marked with a YES. Arguably the most subjective out of all criteria/attributes in these spreadsheets. But also the one I feel strongest about.

    Changing Topics - A paper called “Cues to Deception in Online Chinese Groups” says the following: “we found that deceivers tended to communicate less and showed low complexity and high diversity in their messages.”. I don’t believe in activity tells as much anymore, and complexity is highly person dependent (they can be only compared to themselves). Therefore only high diversity is left, where a slot flip flops between topics and isn't focused.

    Self-pres OMGUS - OMGUS that’s focused on why the accuser is wrong.

    Buddying - Cooperation and Buddying need to be told apart carefully. I expect most slots to be always either one or the other, assuming they play a social game like a social game.

    Spoiler : Town Points :
    Encouraging Content Making new points Providing conclusions Lack of Self-preservation Confirmation Bias Alignment Paranoia Paranoid OMGUS Cooperating
    player1
    player2
    player3
    player4
    player5
    player6
    player7
    player8
    player9
    player10
    player11
    player12
    player13
    player14

    Spoiler : Scum Points :
    Apathy / Coasting Sheeping consensus IIoA (information instead of analysis) Gaslighting / Gish gallop Disassociated Changing Topics Self-pres OMGUS Buddying
    player1
    player2
    player3
    player4
    player5
    player6
    player7
    player8
    player9
    player10
    player11
    player12
    player13
    player14


    I'm not planing to play FM myself any time soon and I don't see bringing myself to go through several games to test them objectively.

    What do you guys think of this sort of scumhunting in general? And do you agree with the Scum and Town attributes chose here to be alignment indicate?
    Last edited by OzyWho; March 23rd, 2021 at 08:48 AM.

  2. ISO #2

    Re: Ozy's Scumhunting Spreadsheets - 2021 edition

    One issue I ran into is that collection of data is itself a time sink and your time/effort is a limited commodity. With that said I have found it to not really be viable to map every post for every player in a game and the times that I have done it I felt like the data I collected was not quite as useful as if I had spent the same amount of time just normally hunting.

    There are things I track such as read walls or posts that are of specific note and I still do go really in depth like your notes suggest sometimes on a player but I have kinda given up on doing it for everyone.

    For your tells I feel like Changing Topics, OMGUS, and Buddying could get you in trouble. Those are kinda context specific tells and also ones that would be more idiosyncratic to the specific player requiring some level of meta thrown in. Confirmation Bias could also be a bit tricky because I have often seen scum players just tunnel scum read 1 player all game.

    In general I feel like you could use something like this to point you in the direction of who to look at though.

  3. ISO #3

    Re: Ozy's Scumhunting Spreadsheets - 2021 edition

    I agree that telling apart buddying from cooperation would be tricky.
    But OMGUS I feel like there are only 2, and both being in the tables is what makes it easy to differentiate.

    Paranoid OMGUS / Tunneling / Confirmation Bias - be they real or fake, it's a trinity of "need to be told apart". I agree that Conf Bias could be troublesome, but it's also one of the things that absolutely must be faked because it's inherently Towny - so not putting something like that in feels like making the whole thing useless.

    Changing Topics I think is one of the safest tells actually, and easy to see in players ISO. But, yeah, context could play a big factor here.


    Ideally, each player would get judged by their own criteria dependant on their weaknesses in their meta. But I want to steer away from meta tells tbh. Like, can I say that I can scumhunt if I can't do it without meta reading the players?


    You're def right that it's time consuming. But I'm thinking that through practice, one could fill such table up more and more efficiently each game whilst playing? Do it 100 times and you can do it in your sleep.
    Then it's only a question of finding a list of criteria that are objective and worth it in the first place.
    Last edited by OzyWho; March 23rd, 2021 at 02:05 PM.

  4. ISO #4

    Re: Ozy's Scumhunting Spreadsheets - 2021 edition

    I did something similar and it was extremely taxing. It was based on player to player interactions.

    My LW from that game:

    My LW and Testament


    https://imgur.com/a/2TtvlIH

    Here is my network analysis of d1 activity. There are 4 charts.

    I reread the whole day and every time I saw a positive/neutral/negative interaction or mentioning of one player to another, or a vote, I counted it. Sometimes an interaction is a single post, sometimes single posts count for multiple interactions.

    For example, post X could have a ren to mm, frinckles, martingg99 interaction generated from it. These interactions are from my perspective and how I interpreted them. I removed fluff, satire, and non-game related interactions.

    The darker the color of the arrow, the more interactions that took place. The arrow starts at a source node, and points to target nodes. So in an example above, there would be three arrows starting at ren and pointing to frinckles, mm, and martingg99.

    I would like to spend some time with you all and see what we can draw from these networks.

    The first chart we will introduce is the "positive interaction/ agreement chart". These occur when it appears to me that two players are in agreement or are otherwise having a positive game related interaction with each other.



    Here you can see, MM and MZ had the strongest positive interactions. What you end up finding though is a few players on the outside. For example, Mesk, Martin, and Frinckles.

    The next chart introduced in the neutral interaction chart. These often occur when one player asks another player a question, or responds with an answer.



    Here you can see the most acive players were MM, DM, MZ, and who they were interacting with.

    This chart is the fun chart, all of the negative interactions, scumreads, scumleans, whatever you want.



    Clearly, there is tension between Mesk and MM, and then a nexus of conflict between DM/SB and Renegade, Marshmallow, and MZ.

    Many others have sat on the sidelines here.

    Finally is the vote chart. This is how many unique times a person voted for someone else:


  5. ISO #5

  6. ISO #6

    Re: Ozy's Scumhunting Spreadsheets - 2021 edition

    Spoiler : :
    Quote Originally Posted by Renegade View Post
    I did something similar and it was extremely taxing. It was based on player to player interactions.

    My LW from that game:

    My LW and Testament


    https://imgur.com/a/2TtvlIH

    Here is my network analysis of d1 activity. There are 4 charts.

    I reread the whole day and every time I saw a positive/neutral/negative interaction or mentioning of one player to another, or a vote, I counted it. Sometimes an interaction is a single post, sometimes single posts count for multiple interactions.

    For example, post X could have a ren to mm, frinckles, martingg99 interaction generated from it. These interactions are from my perspective and how I interpreted them. I removed fluff, satire, and non-game related interactions.

    The darker the color of the arrow, the more interactions that took place. The arrow starts at a source node, and points to target nodes. So in an example above, there would be three arrows starting at ren and pointing to frinckles, mm, and martingg99.

    I would like to spend some time with you all and see what we can draw from these networks.

    The first chart we will introduce is the "positive interaction/ agreement chart". These occur when it appears to me that two players are in agreement or are otherwise having a positive game related interaction with each other.



    Here you can see, MM and MZ had the strongest positive interactions. What you end up finding though is a few players on the outside. For example, Mesk, Martin, and Frinckles.

    The next chart introduced in the neutral interaction chart. These often occur when one player asks another player a question, or responds with an answer.



    Here you can see the most acive players were MM, DM, MZ, and who they were interacting with.

    This chart is the fun chart, all of the negative interactions, scumreads, scumleans, whatever you want.



    Clearly, there is tension between Mesk and MM, and then a nexus of conflict between DM/SB and Renegade, Marshmallow, and MZ.

    Many others have sat on the sidelines here.

    Finally is the vote chart. This is how many unique times a person voted for someone else:

    No conclusions though

    Edit: I think @Frinckles maps "relationship charts" too btw, except for latest reads I think.
    Last edited by OzyWho; March 23rd, 2021 at 01:54 PM.

  7. ISO #7

    Re: Ozy's Scumhunting Spreadsheets - 2021 edition

    Quote Originally Posted by OzyWho View Post
    No conclusions though

    Edit: I think @Frinckles maps "relationship charts" too btw, except for latest reads I think.
    I was too exhausted after going through and labeling everything.

    I think it turned out MZ was scum with Auwt and Mesk, though I thought I was scum with him (strange setup). But hindsight is easy to find in things like this, it is extremely hard to take information like this and make a conclusion.

  8. ISO #8

    Re: Ozy's Scumhunting Spreadsheets - 2021 edition

    There is an entire field of study on that subject called sociometrics.

    "Sociometric explorations reveal the hidden structures that give a group its form: the alliances, the subgroups, the hidden beliefs, the forbidden agendas, the ideological agreements, the "stars" of the show."

    There is a reason I got a little obsessed with this a year or so ago.

    Recently I wanted to create a tool that used sociometrics and combinatorics but I honestly just don't see the data collection part of sociometrics to be viable. Every time I have gone down that road it burns me out so bad I dont even want to look at a mafia game for a while..

  9. ISO #9

    Re: Ozy's Scumhunting Spreadsheets - 2021 edition

    The issue with Renegade's charts is that it either is only quantitative (amount of posts) or qualitative in a very subjective manner. In the first case, information given by the chart is very limited: a single post can be extremely charged for or against a person, or on the contrary, be incredibly weak. In the second case, so many explainations need to be written next to the chart for it to be understandable that it becomes much simpler and much better to simply explain with words. We don't even know whether the author is town or scum, which makes it hard to even trust him. I am not sure it was very useful in that game, either.

    As for Ozy's charts, they suffers from an issue similar to the first one: they do not measure the magnitude of tells, nor do they take into account their context. Each attribute would have to be rated with "weak/medium/strong" or something similar to at least provide some nuance. Once again, many explainations would be required for the charts to be useful and convincing. It's an interesting tool to make sure you've covered every major tell a player could have when you're trying to read that player, but I doubt it's viable as a way to share reads. Sticking to making clear, uniform and relatively concise lists of reads that include the criteria you want to emphasize on is simply better, because it conveys more content while being more accessible and while forcing you to make your conclusions unambiguously clear.
    Quote Originally Posted by The Lawyer View Post
    Besides your lamp and your refridgerators, do you find anyone else suspicious?
    Quote Originally Posted by oliverz144 View Post
    it looks like many, e.g. MM and lag, suffered under the influence of paopan. However there is a victim: frinckles. He left the path of rationality and fully dived into the parallel reality of baby shark, king shark, and soviet union pizzas.
    Spoiler : The meaning of life :

 

 

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •