48

In Close Votes expire too soon for low-traffic tags Jeff exclaimed:

Brock, this is GENIUS! I am hereby using my super-upvote on this feature-request:

Close votes (and probably reopen votes) only expire after all of these conditions are met:

4 days have passed since the last close vote.

The question has more than 100 views.

This has left us with a little bit of a mess:

mess

The review path now has a new a tab that can help you cut through these questions and help us clean up some of the mess. 10k only


I find that having such a huge backlog, makes it very hard to properly deal with closures.

What steps or system changes can we make to get the huge close backlog on Stack Overflow, under control?

Keep in mind, now we keep track of the user ids who "reviewed" the various questions.

13
  • 1
    When doing something about this, would there be a chance for this request to be fulfilled?
    – Pekka
    Nov 11, 2011 at 2:33
  • 2
    was looking at that today ... I need to talk to Jeff about that ... not sure what the reasoning is for not giving you access to deleted close votes.
    – waffles
    Nov 11, 2011 at 2:35
  • 2
    On a side note, it looks like 100 views isn't enough (my anecdotal experience until SEDE lets me prove it). I'm guessing that too many of the "viewers" on stale questions are not able to cast close votes. Nov 11, 2011 at 2:42
  • 1
    I used to like that flagging for closure would also net me flagging weight points. Sadly, I can't do both any longer. Nov 11, 2011 at 5:02
  • 2
    There. I used my 50 close votes for today. Until tomorrow. :-) Nov 11, 2011 at 5:11
  • I'd suggesting pushing the closing privilege to 5,000 rep. Less overzealous people casting stupid close votes. Nov 11, 2011 at 5:40
  • 1
    Are you planning to make it so reviewers don't have to leave the /review page, or should that be a separate request? The new filtering rocks, I just wish reviewing didn't require popping open new tabs and hitting back buttons / refreshing so much.
    – Tim Post
    Nov 11, 2011 at 5:54
  • @tim perhaps post a separate req. I have some ideas for improvements but am open to any suggestions
    – waffles
    Nov 11, 2011 at 6:10
  • I proposed a solution before to a very similar scaling problem, applies here too: meta.stackexchange.com/questions/110404/… Nov 11, 2011 at 7:53
  • This makes me wish the Facebook mini-site moderator tools were filtered so I could help clean that site up. (requested here meta.stackexchange.com/q/103758/143965)
    – bkaid
    Nov 11, 2011 at 16:44
  • @Tim there is a review question link that expands the question in line. doesn't that help?
    – bkaid
    Nov 11, 2011 at 16:49
  • @waffles Could you also add the progress for the "Research Assistant" badge once /review is unlocked? I don't want to open a new question for this since it seems fairly minor (and hopefully, won't need much convincing/discussion) Apr 29, 2012 at 17:52
  • 3
    A year later and the queue is twice the size. I'd love to see something done about this because 48.3k "vote-to-close" questions is absolutely, horribly unmanageable and makes that queue completely meaningless. Let's start with bumping this question with an update about the current state a year later?
    – ErikE
    Jan 1, 2013 at 20:15

8 Answers 8

18

In the same way that popular questions are now harder to delete, perhaps unpopular questions should be easier to close and delete.

1
  • 5
    +1 Demanding 5 votes for C# is wildly unlike 5 votes for sqlite3. The hard number limit has never made sense to me.
    – Andomar
    Nov 21, 2011 at 11:09
13

I don't see the problem.

The way I look at it, there are a couple of reasons why a question only gets one close vote:

  1. The question needs to be closed, but isn't getting much attention from anyone, and so isn't getting much attention from folks who could close it.

  2. The question doesn't need to be closed.

Fortunately, this nifty change handles both cases... 100 people "review" the question, and either 5 of them vote to close, or it doesn't get closed.

Now, there's a potential flaw here in that 100 viewers might not contain enough people who can vote to close to get the job done... If this is actually a problem, then the solution is simple enough: just change that 100 "viewers" to 100 "closers" - users who have the ability to close.

Personally though, I don't think it matters. A question that struggles to get even 100 viewers period isn't exactly poisoning the air on the site - no one's looking at it!

By the way: I really like your new close-review UI. And the ability to filter the list down to questions that have 3 or 4 votes already makes it a snap to jump in and salvage or slam the door on questions that really need it without the 1-vote scatter-shot problem. In other words, I think you've already implemented your solution.

12
  • the intention was never to have 100 closers ... if you look at any question with 100 views probably only 2/3 closers looked at it ... can give you exact stats from haproxy
    – waffles
    Nov 11, 2011 at 2:54
  • should the fact that a "reviewer with close rights" looked at it count as 10 virtual views?
    – waffles
    Nov 11, 2011 at 2:54
  • 4
    If your goal is to simply clear off the "review" list by invalidating close votes faster, sure. But that seems an odd goal, given the point of the previous change was to not invalidate them too quickly.
    – Shog9
    Nov 11, 2011 at 2:58
  • I wonder where this 100 number came from ... why not 150 or 50 or 200? Why age close votes at all? Mainly wondering if we have the right balance here..... What is our goal?
    – waffles
    Nov 11, 2011 at 3:38
  • 3
    The goal is to adjust the vote-expiration time such that low-traffic questions have at least a shot at being reviewed by folks who could close them. Prior to the change, vote-decay was based purely on time. 100 views is a very simple-to-implement attempt at this - I doubt it's based on anything. More precise solutions might involve each view causing votes to "decay" (for instance, 100 views without a vote clears all votes, with each vote resetting the counter), or specifically tracking views from folks with the ability to close, and a past history of closing. Or a combination of both.
    – Shog9
    Nov 11, 2011 at 3:51
  • 1
    If you wanted to come up with a better number, @waffles, finding a correlation between views and close-votes on questions that were eventually closed might be a good place to start.
    – Shog9
    Nov 11, 2011 at 3:53
  • 7
    A substantial amount of the one vote lingering questions follow a pattern. Down votes -> close vote -> one edit -> no more close votes. If no more close votes were received after an edit, the single vote should probably decay much sooner. I'm looking at a lot of on topic, but extremely basic questions where a simple edit stopped the 'way too basic' police from closing something based on quality alone. Granted, I've only gotten through half a dozen pages of them so far.
    – Tim Post
    Nov 11, 2011 at 6:03
  • @waffles: "can give you exact stats from haproxy" this fascinates me, I am tempted to ask a question about what you mean and how you do this, and how I can make my haproxy instance work for me like that! Nov 11, 2011 at 12:52
  • Ah, but "A question that struggles to get even 100 viewers period" does harm the site by polluting the "vote-to-close" review queue (48.3k items right now) and making that review queue almost useless. Over a year later, and we've doubled the number of items in that queue. What's the point of having it, then? I am motivated to help, but what's the point in that queue? No point at all, it seems to me.
    – ErikE
    Jan 1, 2013 at 20:13
  • @ErikE: when the new close vote queue was introduced, there were well over 50 thousand questions in it - it's actually diminished considerably. Since new items are prioritized over old ones, the backlog doesn't seriously hurt anything. That said, it may be worth putting in a hard expiration date for all close votes (regardless of views) just to avoid the perception of augean stables.
    – Shog9
    Jan 1, 2013 at 22:21
  • @Shog9: I just started helping with reviews. during the few weeks the queue grew more then 10K so we are now at 105K entries in the close queue. Something needs to be done. Dec 10, 2013 at 19:02
  • Yup, @Peter - working on it
    – Shog9
    Dec 10, 2013 at 19:22
8

Two suggestions:

  • Allow and encourage tag-filtering of reviewed questions with close votes. That will help individual reviewers to focus on the questions they can best address and are most interested in. This reduces the workload of an individual reviewer down to a much more manageable level.
  • Divvy up the question pool into batches, and assign reviewers different batches, so that a dozen people opening the reviewer console aren't all dealing with the same first 20 questions. This spreads the effort out more evenly among the questions that need attention; it also splits the backlog up into more manageable portions, since reviewers will be able to work at the resolution of a single batch, rather than the entire backlog.
3
  • tag filtering is already there, the batches idea is interesting
    – waffles
    Nov 21, 2011 at 8:44
  • re:tags - Intriguing. Our wee nascent li'l Writers.SE is where I've got reviewing privileges, and there, how shall I put it, we do not yet have this issue. :P
    – Ziv
    Nov 21, 2011 at 9:31
  • @SamSaffron Tag filtering is not really there - you can pick on 3 tags only which is not enough - also being able to filter out for as given tag would be more useful to me
    – mmmmmm
    Apr 12, 2013 at 11:56
6

Don't count close votes on low-view questions more than a month old against the daily limit.

That way the people who are really into site maintenance can clean up as many as they want, without having too much say on the questions people actually care about.

2
  • 1
    are you running out of close votes?
    – waffles
    Nov 11, 2011 at 2:32
  • 3
    Not in months. But if you get the 20 or 50 people who do run out of close votes to cast another 50 or 20 votes per day on old, low-view questions, you'll clear the backlog in a month.
    – agf
    Nov 11, 2011 at 3:56
6

What is the problem to be solved here?

  • The system gets slow when there are lingering close votes. (I don't buy that.)

  • There are questions staying open which actually shouldn't stay open, simply since nobody looks at them to close them.

    If I understand right, this is why we now have this list, so people can find questions which other people also think needs closing, and vote there, too. Having these votes expire sooner doesn't help here.

  • Because of the many questions with close-votes (of which many actually shouldn't be closed), we can't find ones which should really be closed.

    So we need a way to get the ones off the list which should stay open. Either by mark as reviewed (does this work here?), or by some other measure. A moderator could close and reopen them, but a normal user can't. I'm not sure what could be done there.

Do you actual search proposals in which way the software could be changed, or ideas how to get people to go through this list?

2
  • It marks it as reviewed once expanded and never shows it to you again. I am concerned the backlog is too big, that is all. I wonder if the 100 mark is arbitrary ... why not 200 or 50 .... why age close votes at all ... etc.... the post here is part "promotion" of the new feature and call-to-arms. It is also a request for feedback/improvements to this process.
    – waffles
    Nov 11, 2011 at 3:37
  • 7
    @waffles: Aging avoids the problem of a reasonable but long-lived question slowly accumulating enough votes to close it (or re-open it, as the case may be). Imagine if 1 out of every 10 thousand viewers cast a close vote regardless of whether the question needed it or not - given sufficient time, every single question would eventually be closed. We could simplify that by simply killing any question that hasn't been voted on in a while... But of course, that's silly.
    – Shog9
    Nov 11, 2011 at 4:06
5

I think that the new Review/Close tab/function may be enough once word gets around (I didn't know about it).

Tweaks that might help:

  1. Allow "invalid" or "don't close" votes from the review pages. I know this has been asked for, before. (See: "How about a 'Vote not to close'" and "rescind a close vote before it closes" -- both highly upvoted requests.)

    Each such vote would "erase" a close vote, so that it does not show in the mod tools.

  2. Incentivize reviewers more.
    I know we have the Deputy, Marshal, and Reviewer badges, but perhaps a new recurring badge for every 200 (say) votes and/or reviews?

    Or maybe 2 pts rep for every 10 questions that are either closed or had the close-vote zeroed with help from a user. (Capped at, say, 10 pts per day.)

3
  • Um, Deputy and Marshal are badges for flagging. While good, I think (I have no evidence other than hear-say) that we already have a sizeable backlog of flags. Flagging should be a last resort. +1 for the rest of it though.
    – John
    Nov 11, 2011 at 4:06
  • @John, Yeah, you're right as to how they are awarded, but the descriptions say, "Achieved a flag weight... by reviewing and flagging". And while we review, we are encouraged to cast close votes as appropriate. Nov 11, 2011 at 4:12
  • 1
    @John: Users with less than 3000 reputation can't cast close votes, and can only flag questions they think should be closed. Over 3000 reputation, flagging questions is less useful.
    – Peter O.
    Nov 21, 2011 at 9:39
5

Implementing this very popular but status-declined feature request would help.

Can we have the ability to rescind a close vote before it closes?.

At the moment if I vote to close then the issue is resolved then I have no alternative but to leave the close vote hanging there indefinitely.

1

My assumption would be that the ideal situation is that close votes happen organically, i.e. by people browsing the tag in question and seeing a question that needs closing.

The useful ones to catch via review are the ones failing due to low traffic. All the other cases ("nobody else agreed" or "going to happen naturally anyway") handle themselves automatically, without needing to go via /review.

Therefore in the /review route I'd suggest the opposite approach to the current one. Instead of showing close votes sorted by how many they've already attracted or what the reason is (which isn't really the important information) show them sorted according to when they'll expire. Offer "expiring today", "expiring tomorrow" and "this week" (possibly month too).

Within "votes expiring within X days" views you can then rank them by importance, probably number of votes / number of views is a good metric for that since it would put something with 4 close votes from 4 views right at the top, but 1 close vote from 100 would be considered very unimportant.

If you like this then there's a few other things you could consider in the "importance metric":

  • You could possibly use "how negatively scored is the question?" and "how many answers does it have?" as part of the metric, the idea being that very negatively scored or 0-answer questions are possible indicators of close reasons also.
  • There might be some mileage in using "time period over which the currently pending votes were applied" as part of the metric, but I think that's probably needlessly complicated.

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