13

There's been some discussion on meta recently about accepted answer rates. I've added two new graphs to my stackoverflow graphs page:

Accepted answers by day

Accept rate histogram

The first shows how the rate of questions that have accepted answers. Interestingly, the rate took about a 10% drop early on (after SO private beta) and never quite recovered to the same level. The recent steep drop (on the right hand side) shows that the newest questions just haven't had an answer accepted yet.

The second is a user-centric view which shows how many users have each accept percentage accept rate. I used the same criteria as SO itself (more than four questions asked) otherwise the data gets even less interesting.

9
  • The Accept rate distribution does not look that bad. Is it possible to get a histogram for the situation before the Acceptance rate was introduced? Dec 21, 2009 at 20:04
  • I bet the tiny peaks and valleys in "Questions with Accepted Answers by Day" has correlation with the weekends...
    – Tom Ritter
    Dec 21, 2009 at 20:32
  • I assume the date range is from day 0 to 12/21/2009? It's interesting that the %accept feature does not seem to have had an effect.
    – user27414
    Dec 21, 2009 at 20:44
  • The histogram might be more useful with ranges (40-50%, 50-60%, etc). There are spikes at even numbers like 50% (2 questions, 1 accepted), 67% (3 questions, 2 accepted), etc. With ranges that should smooth out and give a clearer picture.
    – user27414
    Dec 21, 2009 at 20:46
  • @John Smithers: It might be possible to create a graph based on when the answer was accepted, versus when the question was asked. I'll give that a go and see whether it's interesting. Dec 21, 2009 at 20:58
  • @Jon B: The date range is day 0 to 30-Nov-2009 (the date of the most recent data dump). I'll also create a histogram with 10% range buckets, and I think a cumulative graph would be useful too. Dec 21, 2009 at 21:00
  • The spike at 0% is a bit sad... I assume (given you said you followed the >4 questions asked criteria) that you're not counting questions with no answers etc?
    – Alconja
    Dec 21, 2009 at 22:06
  • @Alconja: Questions with no answers are counted. Users who have asked no questions are not counted. Dec 21, 2009 at 23:20
  • 1
    @Greg Hewgill: in the first chart, I'm assuming that 75% on (let's say) 2009-07-23 means that 75% of questions asked on that date have accepted answers. Which would explain why there wasn't a big boost when acceptance rate was introduced--people who care would have gone back and accepted answers on their old questions, resulting in a boost across the board. to really see the difference that acceptance rate feature made, you'd need a snapshot of the first graph before and after the acceptance rate was introduced.
    – Kip
    Dec 22, 2009 at 2:45

1 Answer 1

4

Here's a chart of when accepted answers were accepted:

Accepts by day

The obvious spike is the introduction of the accept rate display, and it seems like it may have had a lasting effect. It's sort of hard to tell though.

Also, here's a user accept rate histogram with 10% buckets:

You must log in to answer this question.

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