Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[chaotic skies]

I'm almost waiting for someone to come in and turn this into a learning AI. Then we're boned. Hard.

Also: Updoot plox?

[Flying Dice]

Yep, that's the normal trend. It really throws a wrench into the gears of my prediction algorithms. I should probably try and exclude them from the data set.

I'unno, once we've got enough points to reliably predict new versions it becomes less a problem and more a curious data feature. It wouldn't do to scrub data just because what we do see interferes with what we want to see.

Especially because we can look at stuff beyond new users to get a good idea of how many of those new users stick around.

By the by, your spreadsheets are all down.

[My Name is Immaterial]

(I should have posted this right after I changed the title. )

The Statistics Center is a lie.

At the very top, it claims that this forum has an impressive 3,441,288 posts on it, but when you look deeper into the evidence, a horrible truth is revealed.

There is a massive post coverup.

Do you know how many total posts the top 1% of posters have? 3,804,342. Yeah, that's right.

The top 1% controls 110% of the Forum's posts.

But, if you dig into the secrets they don't want you to find (ie. the raw data), you'll find that the post count is actually a staggering 6,578,298 posts. Where did three million posts go? I don't know.

But I'm 110% sure the Illuminati is behind it.

Here's the raw proof.


Well, that's enough of that.

I'm almost waiting for someone to come in and turn this into a learning AI. Then we're boned. Hard.

Also: Updoot plox?

We won't die in a hail of nuclear fire, we'll die from heart attacks from too many cool graphs.
I don't update until the forum's internal date/time ticks over to February. We've still got about 11 hours left on the clock till then.

Yep, that's the normal trend. It really throws a wrench into the gears of my prediction algorithms. I should probably try and exclude them from the data set.

I'unno, once we've got enough points to reliably predict new versions it becomes less a problem and more a curious data feature. It wouldn't do to scrub data just because what we do see interferes with what we want to see.

It should be easier to just use the release log, instead of the post count to make predictions. I'll try and run up something to figure that out.

Especially because we can look at stuff beyond new users to get a good idea of how many of those new users stick around.

While I would love to figure out that, what should I use as a metric? The closest I can think is Users Online/Day. Maybe comparing the Posts/Day with the UO/D would give me a picture of that.

By the by, your spreadsheets are all down.

Really? Huh. Try this.

[Graknorke]

I am simultaneously disgusted and not at all surprised. This is exactly the kind of thing we'd expect from the capitalist wasteland of Bay12.

[TheBiggerFish]

I think it's more that SMF might just be unable to count.

[Zanzetkuken The Great]

Wonder if subboards, Armok 1 and the various nonsense would account for the discrepancy.

[Flying Dice]

There's likely some error due to hidden/locked boards and such. Maybe deleted threads/posts as well, I'm not sure how SMF stats handles those.

That's what I was thinking. Work out the new registered users/month for a given period of time--I'd probably count the 6mo after a release as being a release spike, since that roughly matches the spikes in registration. Then look at the average unique users online for the month prior to the period you're examining, compare that to the AUUO (say) one month after the end of a release spike. That'd probably give a decent read on how many regular users the forum picks up from each release. Maybe get your best numbers by comparing from 1mo prior to a given release against 1mo prior to the release after the first one, to get as close as possible to discounting people who eventually wander away.

Granted it's pretty much impossible to get a pinpoint accurate estimate on any of these various extrapolated statistics because of the sheer number of variables involved.

[My Name is Immaterial]

I think it's more that SMF might just be unable to count.

That's just what they want you to think.

Wonder if subboards, Armok 1 and the various nonsense would account for the discrepancy.

[srs]I'm not seeing all of the subboards adding up to 3 million, but if you could check, I'd appreciate it.[/srs]

snip

I'll take a crack at it in a bit.

[Loud Whispers]

No no, it's fine, the post counts will trickle down to the escaped lunatics

[Shadowlord]

What is this "statistics center" of which you speak?

[TheBiggerFish]

The thing at the bottom of the top of the boards.


http://www.bay12forums.com/smf/index.php

[wierd]

Another potential source of the error is moderator deleted posts.

If the counter increments when they are posted, but does not de-increment when they are culled, then the values will quickly diverge.

If that is the case though, the descrepency would itself be an interesting data point.

[TheBiggerFish]

That doesn't add up.

50% of all posts having been deleted just doesn't make sense.

We must test this theory...

[Glloyd]

tfw not in the top 1%.

Gotta start shitposting.

[TheBiggerFish]

tfw not in the top 1%.

Gotta start shitposting.

Good luck, dude.