53

End-of-year posts seem to have become the tradition these days, so I'm back again with a third Year in Spam. Welcome back, folks, to another round of statistics you never needed to know.

All times/dates are UTC.

The Big Number

In total, we saw 35,636 spam posts this year across the network. That's down from last year:

 Year | Posts | % change
------+-------+----------
 2016 | 31578 | 
 2017 | 39656 | +25.6%
 2018 | 35636 | -10.1%

Posts By Site

Ask Ubuntu continues to catch up with Stack Overflow this year - from a difference of just under 7% last year it's now down to just over 4% behind. Other sites in the top 20 remain in fairly similar places, although The Workplace has dropped significantly.

+------------------------------------------+------------+------------+
| Site Name                                | Post Count | % of Total |
+------------------------------------------+------------+------------+
| Stack Overflow                           |       6892 | 17.3795%   |
| Ask Ubuntu                               |       5237 | 13.2061%   |
| Super User                               |       3294 | 8.3064%    |
| Drupal Answers                           |       2058 | 5.1896%    |
| Ask Different                            |       1853 | 4.6727%    |
| Graphic Design                           |       1687 | 4.2541%    |
| Meta Stack Exchange                      |       1468 | 3.7018%    |
| Astronomy                                |        620 | 1.5634%    |
| English Language & Usage                 |        600 | 1.5130%    |
| Bitcoin                                  |        561 | 1.4147%    |
| Role-playing Games                       |        559 | 1.4096%    |
| WordPress Development                    |        483 | 1.2180%    |
| Puzzling                                 |        450 | 1.1348%    |
| Web Applications                         |        437 | 1.1020%    |
| Travel                                   |        416 | 1.0490%    |
| The Workplace                            |        362 | 0.9129%    |
| Arqade                                   |        324 | 0.8170%    |
| Android Enthusiasts                      |        323 | 0.8145%    |
| Personal Finance & Money                 |        320 | 0.8069%    |
| Mathematics                              |        309 | 0.7792%    |
+------------------------------------------+------------+------------+

Truncated to top 20 sites. View full data.

Posts By Time

2018 has mostly followed 2017's line in this regard, though spam has been a little more spread out through the day (i.e. less confined to peak hours). Behold - the chart returns (the y axis represents the proportion of spam posts created in each hour group).

enter image description here

Time to Deletion

2018 was not a great year in this regard — despite (or perhaps because of - we're not really sure1) improvements to the autoflagging system. The time it takes, on average, to delete a spam post, was higher at all hours of the day than it was in 2017 (although there are still some gains over 2016).

enter image description here

Speaking of autoflagging...

Autoflagging

Charcoal's autoflagging project currently stands at 45,183 posts flagged, of which 44,990 were confirmed to be spam - that's 99.57% accurate overall. There are 335 users participating - 31 of those users are moderators, of 173 36 sites between them (bad Catija, skewing my numbers with CM multi-diamonds).

This graph shows the overall progress of the project: a number of posts flagged and flags cast per month, since inception.

enter image description here


1 This could also be down to improvements in TTD tracking this year, which remove a previous minor bias towards lower values.

  • 11
    Oops! Sorry? 😀 – Catija Jan 2 at 2:15
11

Thank you for sharing in public what the effort you and all that participate in several roles have achieved and how it helps in making Stack Exchange sites a better place.

I do know most Charcoal visitors use userscripts to be more effective with handling spam posts. Do you have stats on which means are used by those users you can keep track of? If you don't have those stats, is it worth adding some measurement so you can report on those over 2019?

If we still believe that expedite removal of spam is warranted, which measures and/or actions are needed by us flaggers to achieve better time to deletes over the whole day.

I have have been contacted twice or thrice over the last year where Smoke Detector panicked and informed me of an auto-flag gone south, mostly leading to a declined flag in my flag-history . How many of such cases in 2018 did we have?

  • w.r.t the last point, there is an open feature request that would completly stamp out the issue – Magisch Jan 2 at 9:39
  • AFAIK all our userscripts contact the metasmoke API, without interacting with SE directly, so "spam flags cast manually" includes ones cast via the metasmoke API. This also means that we're in control of the usage of userscripts. – ɪBᴜɢ Jan 2 at 9:39
  • On your last point, check out my Autoflag FP dashboard. According to that, last year we had 85 fps on autoflagged posts, out of a total of 19,464. – CalvT Jan 2 at 12:08
  • 3
    I threw some data about flags cast through userscripts into the post. – ArtOfCode Jan 2 at 15:44

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .