15

While discussing the effect of a beta site, attention turned to accept rates on SO. There's a sharp fall in global accept rate (not per-user) in early 2015:

enter image description here

This is also true of Ask Ubuntu:

enter image description here

And Super User:

enter image description here

And of Server Fault, and to a less degree, Unix & Linux, but not TeX - LaTeX or English Language & Usage.

What happened? Or did nothing happen, and I'm reading too much into happenstance?

  • 2
    Love the 2015.0 in thi image. – BRHSM Feb 12 '16 at 21:42
  • @CoderGuy that's Gilles going for a stardate feel. ;) – muru Feb 12 '16 at 21:43
  • 3
    The lack of a vertical axis labeling in these graphs makes me sad. – Nathan Tuggy Feb 12 '16 at 22:15
  • @NathanTuggy, the lack of vertical axis labeling makes the answer make no sense to me at all. If a bunch of non-answered questions (which therefore don't have accepted answers) are deleted, wouldn't that increase the ratio of "Qs with accepted As"/"Qs without accepted As"? What am I missing? – Wildcard Apr 4 '16 at 20:29
  • Also, xkcd.com/833 – Wildcard Apr 4 '16 at 20:30
  • @Wildcard: That doesn't have much to do with the vertical axis, but the horizontal. Precisely what you describe is precisely what is happening in a rolling window moving forward from the past. That's why the ratio appears to fall off at a particular point. – Nathan Tuggy Apr 4 '16 at 20:36
  • @NathanTuggy, I would expect to see a spike rather than a drop in the rate following deletion of unanswered questions. What do you mean by "rolling window"? – Wildcard Apr 4 '16 at 20:44
  • @Wildcard: That's what we do see, because deletion happens at a point in the past that keeps moving forward and moving the spike forward with it! The "window" between posting and automatic deletion is of a fixed length of time that rolls forward continuously. – Nathan Tuggy Apr 4 '16 at 20:48
  • @NathanTuggy, I'm truly trying to understand your comments. I understand the SQL linked from the question. The phrase "a point in the past that keeps moving forward" is ill-defined, as is "a fixed length of time that rolls forward continuously." Do you mean that as time proceeds, new deletions occur, so the "date of most recent deletion" is not a static value but jumps forward with each new deletion performed? Do you mean that due to the set interval for which unanswered questions remain undeleted, at any moment there are some questions which will soon be deleted as they reach sufficient age? – Wildcard Apr 4 '16 at 21:21
  • @Wildcard: Both. But the crucial point is that that whole plateau on the left is the past, where deletions have already been processed and the ratio is therefore the highest; the cliff near the right is the roomba's current progression, and the valley extending to the rightmost is un-roomba'd questions, containing many that are unanswered that drag the ratio down. And that the cliff moves to keep the valley at a (roughly) fixed width. – Nathan Tuggy Apr 4 '16 at 21:26
  • @NathanTuggy, aha, thank you. That makes perfect sense now. For some very stupid reason I was thinking the deletions occurred once per year, from which standpoint your comments made zero sense. Silly of me; thanks for your patience. :) (Also I can see now that the very steep drop at the very end of each graph is for similar reasons—there are always questions asked in the current month which will be deleted by the following month.) – Wildcard Apr 4 '16 at 21:31
  • Also I had imagined the SQL being run each month, over time—the data point for each month having been generated/reported at the time rather than now based on the current set of open questions that were simply created in the past month. From this standpoint you can see the spike would show in the other direction on the graph. Basic assumptions in error; always obvious in retrospect.... :) – Wildcard Apr 4 '16 at 21:33
  • @Wildcard: No problem; it took a bit of thought to wrap my head around it as well. – Nathan Tuggy Apr 4 '16 at 21:41
18

Its a year ago. It is an artifact of the Roomba script frequency.

There is a Roomba script that deletes questions that don't have an answer (among other requirements) and are older than 365 days. Deleting these questions will cause the ratio of questions with an accepted answer to questions that don't have an accepted answer to shift.

  • Ah, so the fall will keep shifting over time. – muru Feb 12 '16 at 20:20
  • 1
    Yep. Look at the 365 day script and consider what happens when those posts are deleted. – user213963 Feb 12 '16 at 20:24

You must log in to answer this question.

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