One quick side note on this, it would be a lot easier if the almighty @DRow would be so kind as to gift me with an API key.
I assume we are considering scraping birthday data as cheating?
That is a clever move, but user provided data cannot be trusted. You have to use trustworthy system generated data.
What I had in mind for you is to look at something like the average time users spent reading per visited day over the last quarter, number of visits per months, number of posts and topics read per previous month vs quarter or year, how many posts they read before posting one themselves, estimated time spent in PM chats vs open side of the forum, etcā¦
This will let you group users by several distinct forum interaction behaviors. Those individual criteria should be relatively intuitive to figure out even without neural net, but any one of them alone is very unlikely to give you a clear answer about the age of the user.
Where the power of the neural network comes in - is the ability to detect a hidden pattern in the unique combination of those distinct behavioral groups that could hint at the age of the user.
I love that @Foster is the pinnacle of coachness 
whats my score @Sylvie
I think there are some more, maybe more cheat-y and inaccessible data points that would determine pretty decisively whether one is a coach or a student.
- Average number of capital words and periods per post.
- Number of likes given.
- Whether ācoachā or āmentorā is a part of the username.
- Average reply length.
- Average likes per posted topic.
- Number of flagged posts.
- Average or median PM thread length.
- Number of image posts in āDankest Vex Memesā
- Number of likes given and received specifically in āDankest Vex Memesā.
- Whether their profile pic is a letter with a solid color background or not.
- If their name is Foster, theyāre probably a coach.
Thereās probably more, but those are the ones off the top of my head.
Lol, sorting everyone on a random subjective criteria is fraught with making many people unhappy.
We were quite infuriated at @Sylvie for attempted willful deceptive list abbreviation and calling us āa fewā among many other things he is guilty of!
You donāt just mess with those who are on the top of the āusers with no lifeā leaderboard!
This is really cool, Taran. I applaud you for doing it.
This was included in the algorithm
Is this data available?
I donāt think I can get this. That being said I know where I fall (lol). 75% of my activity is in DMs.
If Iām gonna go down this path, Itās gonna be aā¦ long project. A neural network is not something Iāve worked on before, and I really donāt know where to start. I think I have a method for working with the data without one, but weāll see how well I can tune this.
Yeah, that surprisedā¦ no one.
Uh Iām not at home rn, and Iāll need to sleep when I get there but I can run my script again tomorrow for anyone who wants it.
Possible to get
Possible to get
Surprisingly, likes given per minute spent on the forum is the only data field I have without a trend in either direction.
Likes per post is available, but it doesnāt distinguish between replies and topics. I can say that @Illyana manages an average of 10 likes per post, well ahead of anyone else.
Unavailable to me.
Unavailable to me.
Unavailable to me
Uh, possible to find, but like, would take ages to download and parse all the data.
The man has a point.
I take this back, here you are!
whats my score @Sylvie
Here, I ran the algorithm on the top 1K users.
This first version of my algorithm works like this:
- Every data field = Every data field/time read
- Take the mean of every data field
- Every data field = (Every data field/mean) - 1
- Sum all data fields
This works somewhat because I noticed that coaches tend to have generally-higher stats in every field per time read.
I will note that the numbers come out different depending on who all is in the data set.
So mk. 2 looked like it was a lot more accurateā¦ but then I saw this at the top of this list. Back to the drawing board!
yeah foster not being at the top is totally off
/illegal 20 characters
Speaking of code and data, I was thinking of trying to build a vexibot-style bot for this forum that would give team and user statistics on command. I certainly would know how to go about doing this if I were to try it.
Itās simply a matter of if people would want it, and if the admins would let me do it. I have previously built a successful program to bridge a private message chat and a Discord channel.
I like it. Sounds like a cool idea. Kind of like Mee6.
I think it would be useful, although it would get annoying if people start using it too frequently and creating clutter and spam like what happened with @discobot.