2
$\begingroup$

When I was answering the 'ridiculous question' question, I got to wondering about whether or not one could use the count of answers provided as some indication of the relative quality of borderline questions. I'd like to ask here though about going the other way.

To that end, I'm interested in knowing if there is any correlation between question score and answer count. If there is, are there questions with an unusually low score by comparison to the count of answers that they have received. If there is, and they're still open, has anyone considered the idea of adding bonus rep points to a user's rep (instead of the question) if the ratio falls below a certain ratio, the question is still in positive reputation territory and open?

At the very least, it would seem to be worth a badge of some kind.

I want to stress I'm not personally advocating this; it's a legitimate call for thoughts on whether or not attracting a high volume of (particularly high scoring) answers would be worth something to a question poster.

$\endgroup$
4
  • 1
    $\begingroup$ Kingledion has created some cool analysis in his Holiday Data-palooza: Site statistics from 2017, but not quite what you are looking for (question score and answer count). There are some questions with a low score and a high amount of answers. Questions about torture sometimes fall into that category, or other difficult topics. Here is a closed one, though you are asking for an open one. $\endgroup$
    – Secespitus
    Feb 24, 2018 at 9:34
  • 1
    $\begingroup$ Also see How strong could a dirty mercenary army be? with +8/-8 and How effective would an army of succubi have been in medieval times?. I am not sure what you mean about the user's rep instead of question rep. A bounty for question? If there are questions like the linked ones that generate a lot of up- and downvotes I am not sure if we really want to encourage that with a badge or anything like that... We should encourage good answers not discussion generating questions. $\endgroup$
    – Secespitus
    Feb 24, 2018 at 9:38
  • $\begingroup$ Both good points @Secespitus - Some interesting questions there and I have to agree with you about the large up and down votes. Hadn't thought of that. Interesting about the torture tag; didn't even know we had one of those but I can see why a lot of people would find that difficult to upvote. $\endgroup$
    – Tim B II
    Feb 24, 2018 at 11:20
  • 1
    $\begingroup$ Net score per non-deleted question, and all non-deleted answers per question (thus giving you a count of answers), are available via the Stack Exchange Data Explorer (SEDE). It would certainly be possible to put together a query that correlates the two, and the graph data visualization mode should tell you if there's any significant correlation (even if just for a part of the span). $\endgroup$
    – user
    Feb 24, 2018 at 12:39

1 Answer 1

6
$\begingroup$

In general, answer count correlates with score for all posts (both closed and open). While closed questions are generally lower scoring considering the number of answers they get, the difference is not big enough in my opinion for it to be valuable as any sort of quality indicator.


If you have any questions, feel free to ask.


I used the following code in SEDE to get a CSV file of all the questions:

select answercount, count(*) as [heat], score from posts 
where parentid is null and answercount is not null
group by answercount, score

I used the CSV and an online plotting tool to this heatmap (note that the big weird boxes on the upper right of these graphs are all single points... I'm not sure why it displays like that):

Here's the query to get a CSV of all non-closed questions:

select answercount, count(*) as [heat], score from posts 
where parentid is null and answercount is not null and closeddate is null
group by answercount, score

Heatmap:

And closed:

select answercount, count(*) as [heat], score from posts 
where parentid is null and answercount is not null and closeddate is not null
group by answercount, score

Heatmap:


Note: These queries can be changed to work better (but not great) in SEDE. Just change the first line of the query to

select answercount, score from posts 

And also remove the last line.


$\endgroup$

You must log in to answer this question.

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