I'd like to know the percentage of questions that are deleted by Stack Exchange's automated deletion scripts, which later get undeleted, on each site.

Initially, I thought I could query for this in the Data Explorer, but it turns out that questions that have been deleted by the Roomba which get undeleted later, will be immediately re-deleted (once the script runs) if the criteria still apply to them. As the Data Explorer excludes deleted questions, it would be nice if someone on the inside could query the database for this.

Ideally, I'd like the results to be in a row for each site, including meta sites. Each column should include:

  • the percentage, out of all questions that were deleted by the Roomba scripts (preferably including those deleted before those deletions were attributed to Community), of questions that were undeleted at least once after being deleted by the Roomba (including ones that were deleted again)
  • said percentage broken down by each respective Roomba criterion (e.g. out of all questions deleted by RemoveAbandonedQuestions, the percentage of such questions which were undeleted later, and repeat for all criteria)
  • for each of the above, the percentage of those questions which were later re-deleted again, preferably broken down into two statistics: those deleted by the same script, and those deleted for a different reason (a different Roomba script, humans voting to delete, owner's account being deleted, etc.)

Based on what I've read in many posts on meta, I'm expecting the first percentage above to be rather small. However, I'd like to concretely know how small it is, or whether or not it's bigger than what's expected.

  • 3
    What about the questions deleted by Roomba that instead of being undeleted were reposted ?
    – Rubén
    Dec 15, 2020 at 3:00
  • @Rubén A Stack Exchange employee commented on another question that there's no way to programmatically detect those. Dec 15, 2020 at 4:50
  • If I understood correctly the employee says that is not possible to repost same question (same title and body and the one deleted by Roomba)... I just found a post like that a couple of days ago.
    – Rubén
    Dec 15, 2020 at 5:24
  • 2
    @Rubén As I said, they said that there's no way to programmatically detect reworded reposts. What probably happened in that case is that the author changed one word (or character) to have their repost be allowed. Dec 15, 2020 at 5:28
  • Checking the markdown they are exactly the same, character by character... actually there are three exact copies, two deleted and one posted few days ago
    – Rubén
    Dec 15, 2020 at 5:40
  • 1
    @Rubén Could be possible that the author posted it with different content, then edited it to be the same during the 5-minute grace period. Dec 15, 2020 at 5:45
  • Is there any way that we (or a diamond mod) can be certain of that?
    – Rubén
    Dec 15, 2020 at 5:46
  • 1
    @Rubén 0 point in changing the question if the question is on-topic. You are wasting the OP's time. Dec 15, 2020 at 7:28
  • 1
    somewhat related: meta.stackoverflow.com/a/403095/578411
    – rene
    Dec 15, 2020 at 8:05

1 Answer 1


Since the 2016 question 1-year-old questions removed by Roomba: what happened if they are re-asked? never got any response, I'll share my own experience with my own questions.

I've had between 200 and 300 questions removed by Roomba, spread across >10 SE websites (I don't think I can compute my exact number of questions removed by Roomba unless I go through the dumps). Amongst them, 0 were undeleted. I haven't voted to undelete any of them because as you mentioned, undeleting a Roombaed question is quite useless since it'll get removed again by Roomba within a few days (unless the question somehow attracts some votes/comments/answers, which is unlikely since undeleting doesn't cause the question to be seen on the first page).

I believe if someone wanted to go through the undelete process to salvage their questions, then they should make some random edit to the question right after it gets undelete so then the question appears as active and perhaps with some luck attract some votes/comments/answers. Sounds like more hassle than simply re-posting the question.

Of course all this is ridiculous. The rules for Roomba are badly chosen. E.g. see decrease the required view count to prevent Roomba deletion when score = 0 on smaller sites or 39 questions of mine deleted, and counting. How do you expect users to write well written questions?. Roomba is especially detrimental to expert questions (since the most advanced a question is, the fewer views+answers it'll get), which means that Roomba is discouraging experts from asking questions on SE.

  • 2
    Thanks for the opinion, but this isn't an answer to the question I've asked, which is a concrete request for network statistics. Dec 15, 2020 at 7:26
  • 2
    @SonictheK-DayHedgehog This is an approximation based on a decent sample size. Better than nothing, still valuable. Dec 15, 2020 at 7:27
  • 2
    I completely understand what you're saying. The point of this statistics request is because I'm planning to later make a feature request, to "reset the clock" in cases where a Roomba'd question is later undeleted. Statistics here will help provide backing for (or against) such a request. I'd like to make a concrete case which makes it clear that a really small number of Roomba'd questions get undeleted anyway, so implementing the request wouldn't leave a litter of bad questions behind (appeasing those who have the opposite, favorable POV on the Roomba). Dec 15, 2020 at 7:58
  • 1
    When one has concrete statistics based in fact, it (at least here on SE) results in people from both sides listening, while not having them makes a weaker case. Dec 15, 2020 at 8:01
  • 1
    @SonictheK-DayHedgehog understood, I'd be interested in this stats too but SE employees didn't answer my related question, so I doubt they'll answer this one, hence my approximation. Dec 15, 2020 at 8:26

You must log in to answer this question.

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