14

Since the recent change to Close-Votes the number of votes in the Close review queue has been growing dramatically. This has been discussed. Most of these discussions center around whether there is any hope of reducing the queue to zero, like the other review queues, or somehow increasing the capacity to work through the backlog so it's at least stable. While I agree with these general desires I also want to have a better understanding as to what is creating the situation.

As I see it:

  • A question gets asked
  • Some number of them are marked as closed and enters the queue.
  • Some number of those are reviewed four more times and either closed or left open, and the question leaves the queue.

What I don't understand is whether the ratio of the second point over the first point has changed, or the third over the first or second. These metrics are more important to understanding the system health than the absolute number of questions in the close vote queue:

  • Are there more bad questions? (third over first)
  • Are questions being marked to close more readily? (second over first)
  • Are questions being marked to close more often than they're reviewed for close? (third over second)

Is it possible to get access to these numbers somehow? I think that they could inform the discussion on what to do (or if anything need be done) about the close-vote review queue. (For instance, it may be appropriate to disentangle when you can vote for something to be closed from when you can review something to be clsoed.)

3
  • You are forgetting what may also be a large aspect of the close votes queue: old questions. What is on topic on stackoverflow has changed drastically over the years so there are many old questions that people vote to close because they no longer follow our standards.
    – Josh Mein
    Aug 23, 2013 at 18:26
  • 3
    As of today, there are 71k votes. Earlier this year that was down in the high 20s/low 30s. Therefore the majority of the votes must be from the current funnel, and not the backlog. Further, this is actually addressing exactly that issue; by wrapping metrics around how the funnel is shaped rather than just taking one raw number that includes things such as our backlog. Aug 23, 2013 at 18:49
  • Although most of the close votes are most likely recent ones, the somewhat recent changes to close votes have added to what some people would think should be closed; therefore increasing what questions could be closed. Honestly, I think the real issue is that people are not searching before they enter questions. I would like to know how many questions are currently being voted to close as being duplicate.
    – Josh Mein
    Aug 23, 2013 at 18:55

2 Answers 2

6

The reason is that now flagging question for closing from users with under 3k reputation no longer reaches the flag queue, but the close queue.

It's mentioned here.

So it's not that there are more bad questions, it's just that there are more "events" that can put a question into the close queue now, and no increase in the number of reviewers.

7
  • 2
    Where are the metrics to back this up?
    – Cole Tobin
    Aug 24, 2013 at 5:02
  • @ColeJohnson: What metrics are you looking for, exactly? Aug 24, 2013 at 5:19
  • 1
    Like a graph showing that when that was changed, the number rose exponentially
    – Cole Tobin
    Aug 24, 2013 at 5:20
  • @ColeJohnson: Only a Stack Exchange employee would be able to compile such a graph. However, it's clear from the times of the changes that they are closely related. Aug 24, 2013 at 5:25
  • 2
    While I agree that this is definitely a new source of close-queue votes, I don't agree that not knowing the metrics defined in the question isn't important. I don't see how this answer directly responds to the question. Aug 24, 2013 at 8:38
  • 1
    @NathanielFord I may just have to write a cron script that queries the page and gets the number every few hours. Then I could use Google's graphing "API" to make a graph. That might help. Maybe web.archive.org could be used to gather numbers for past dates?
    – Cole Tobin
    Aug 24, 2013 at 20:52
  • @MadaraUchiha - I believe there is sufficient data at data.stackexchange.com to build a graph.
    – JDB
    Aug 28, 2013 at 17:42
2

It sounds like you are looking for the Stack Exchange Data Explorer.

For example, I ran this query and was able to determine that for all close votes cast in May, 2013 (the latest full month of data available):

  • 91.39% of close votes were cast on posts created in the same month
  • 96.87% of close votes were cast on posts created in the same year
  • 2.33% of close votes were cast on posts created in 2012
  • 0.61% of close votes were cast on posts created in 2011
  • 0.14% of close votes were cast on posts created in 2010
  • 0.04% of close votes were cast on posts created in 2009

These statistics are simplistic. For example, they look only at votes, not "first votes" or flags (which is what would cause something to appear in the queue). You would need a more sophisticated query to extract just the first vote on each post (which I may get around to writing some day).

You can download your query results and run them through the statistical software of your choice if you are interested in performing data-mining, etc.

You must log in to answer this question.

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