70

Last year, I posted 2016: a year in spam. Since that was pretty well-received, here’s the 2017 version. Ladies, gentlemen, aliens, and everyone in between: more statistics about spam that you never needed to know.

Caveat/disclaimer: this data is from metasmoke, the web dashboard for SmokeDetector. It picks up a very large proportion of spam that hits the network, but there may be a post or two here and there that isn't included in the data.

All times/dates are UTC.

The Big Number

Network-wide, we saw 47,841 spam posts in 2017. That’s up 47.3% from 32,462 in 2016. Some of that is attributable to improvements in detection techniques, but there is a general upward trend in spam hitting the network.

Posts By Site

Stack Overflow saw less of the total spam this year than last year — down from 26% to 21%. Ask Ubuntu, Ask Different, and Graphic Design are all up, however, while Super User and Meta Stack Exchange continue to drop.

+------------------------------------------+------------+------------+
| Site Name                                | Post Count | % of Total |
+------------------------------------------+------------+------------+
| Stack Overflow                           |      10220 | 21.3624%   |
| Ask Ubuntu                               |       7058 | 14.7530%   |
| Ask Different                            |       4353 | 9.0989%    |
| Graphic Design                           |       4184 | 8.7456%    |
| Super User                               |       3336 | 6.9731%    |
| Drupal Answers                           |       2840 | 5.9363%    |
| The Workplace                            |       1521 | 3.1793%    |
| Meta Stack Exchange                      |       1242 | 2.5961%    |
| Astronomy                                |        794 | 1.6597%    |
| Information Security                     |        639 | 1.3357%    |
| Arqade                                   |        500 | 1.0451%    |
| English Language & Usage                 |        458 | 0.9573%    |
| Mathematics                              |        424 | 0.8863%    |
| Android Enthusiasts                      |        419 | 0.8758%    |
| Web Applications                         |        410 | 0.8570%    |
| Personal Finance & Money                 |        397 | 0.8298%    |
| Travel                                   |        345 | 0.7211%    |
| Bitcoin                                  |        307 | 0.6417%    |
| Server Fault                             |        292 | 0.6104%    |
| WordPress Development                    |        267 | 0.5581%    |
+------------------------------------------+------------+------------+

Truncated to top 20 sites. View full data.

Posts By Time

The peak spam hours have changed a bit since last year. The peak now starts later and lasts a little longer, but the core hours have stayed the same. Behold, a chart!

Most spam is posted between 04:00 and 11:00.

Time to Deletion

It took less time to delete a spam post in 2017 than it did in 2016. On average, deletion came just over a minute quicker this year — 69 seconds faster.

Spam takes a little longer to get deleted when more of it is posted.

This is in part down to the autoflagging system that Charcoal has been running this year. Speaking of which...

Autoflagging

Okaaaay, I’m going to cheat a bit here and include all autoflags (that’s from December 26 2016) rather than just those in 2017. 26,002 posts have been autoflagged, totalling 68,252 flags. 25,930 of those were true positives (correct detections — i.e. spam posts) — that’s 99.72% accurate overall.

The number of posts we’ve flagged through the year has stayed pretty consistent each month. The number of flags cast, however, has changed as we’ve adjusted thresholds and the maximum flags cast.

We started with one flag per post, but now we do three flags per post.

As of today, 208 users are participating in autoflagging, 23 of whom are moderators (of 24 sites between them). Our autoflaggers have also cast 33,528 flags manually through Charcoal’s various tools.


The data for the charts in this post, including the charts themselves, are available in a Google sheet.

  • 22
    One note: The posts that Smoke Detector and metasmoke see and flag are after posts have gotten through Stack Exchange's quality checks and SpamRam. So, the ~48K spam posts seen last year does not include anything blocked by Stack Exchange. – Andy Jan 3 '18 at 18:08

You must log in to answer this question.

Browse other questions tagged .