Are there software tools for analysing forum threads?

[Cabd]

So, iirc Psyche has an API to do this specifically for MS. I've been debating ingesting it into a BI tool and doing some analytics.

[Cabd]

Software like that should be seen as a great equalizer.

[Cabd]

Equalizer of what?

Time.

I used to keep a literal metabinder of things that majorly differed between an individual's town games and mafia games, because I had the free time to read almost every game onsite.

Turns of phrase, post count, post density, post timing... there were a LOT of super-specific tells like that.


But now? I'm a father and I work full time and I don't have the time to do that. Automating part of that process, and then applying human logic to the patterns it detects? That could be incredibly useful.

And it'll march the arms race along further. But, you cannot stuff the genie back in the bottle. The working solution to countries obtaining nuclear weapons was never "That's unfair, you should stop doing that"; just as it was for the shotgun (Famously detested by the Germans during Trench Warfare) or the longbow, or the battering ram, or any other weapon in human history.

[Cabd]

Essentially, there's two "looks" for how this can go.

Look A: Players like myself and psyche develop our tools, and they stay with us, giving us an advantage

Look B: Players like myself and psyche develop our tools, post them for public use, and everyone advances their games.

[Cabd]

For instance, right now, I'm working on a PowerBI desktop (Free tool from Microsoft) script that ingests a game's worth of posts, and mines them for phrases and words of a given alignment. I tell the model who is town vs scum, and it tells me what words pop out to it.

[Cabd]

People are already fed up with meta usage

Kinda don't think you can speak for everyone here.

[Cabd]

The new parts of the site that were in the works included functionality to do a secret alt each game... but don't expect that to ever materialize.


You can always create a secret alt each game, but i doubt the player base writ large will do so.

[Cabd]

I hope you didn't mean me by "People who don't have any idea how this would work" because uh.... I already built the Alpha for the "word cloud" tool?

Spoiler: Alpha Screenshots

Here's a completed newbie game (40ish) pages. This currently has to manually scrape and load each post, but admittedly will be much faster once the API is available.

Here's the completed game thread:
viewtopic.php?f=11&t=84378

Here's what the tool sees when I load the first 200 posts (AKA 8 Pages), filtered by town posters and "repetitions" set to 5 or above:


And for mafia with the same settings...



Now... tell me what about this is at all somehow different than a player going through the first 8 pages while noting phrases they see?

[Cabd]

But like, Cabd, what if you told it to compare the same player in different games of each alignment?

Well, here's Morph the cat in two Micro games. One town, one mafia....

Spoiler: Mafia in Micro 231

Spoiler: Town in Micro 719

Does this data make you feel like you can solve morph's alignment in a third game with a very high rate of success?

[Cabd]

I should mention all of this analysis is being done using the free Power BI tool from Microsoft: https://powerbi.microsoft.com/en-us/desktop/

Using the "Web source" function and the "Word Cloud 2.0" visual.

And a bunch of Powerquery M under the hood to strip the HTML tags and assign alignments.

[Cabd]

Do you think MORE game data put in to these models would have helped?

[Cabd]

Well, once the API releases and doing this doesn't kill the site with traffic, I'll release the source code publicly and people can try it out themselves.

I don't think it's going to change much!