« first day (1640 days earlier)      last day (1778 days later) » 

6:33 AM
@MattSamuel could you quantify your claims regarding mass deletions? If a serious debate is to be had this seems important. Say, starting from which percentage do you think it is mass deletion? — quid ♦ yesterday
@quid Quantify? What do I look like, a mathematician? In seriousness, I can't attest to the quantity being large compared to the total number of questions. But I consider hundreds to be mass deletions, and I'd estimate given my anecdotal observations that it's at least of that order. — Matt Samuel yesterday
@MattSamuel "hundreds" over what period of time? In total? Per day? What else? Assuming total, given that there is more than a million questions on the site "hundreds" will be less than 1 in thousand. If that is your opinion I encourage you to be up-front about this and to frame your critique in that way. It would also mean that you seem to think that there is not even one question per day newly asked that should be deleted. Do you actually think this? — quid ♦ yesterday
@quid I'm focused mostly on the deletion of old questions. — Matt Samuel yesterday
@MattSamuel define "old". I mean it ought to be possible for you to articulate a precise objection. And if you cannot then it would seem apt to present it as such, as some vague sentiment that you are unable to make precise. There is an actual risk in users believing all kinds of stuff just because other users repeat something all the time. Mass deletions of old questions happen. No one knows how many, no one knows what old means. But certainly it sounds scary. ;-) — quid ♦ yesterday
One could play with parameters - I looked at three years and minimal score 5.
And of course, many various SEDE queries could be made to get some data. (In particular, if somebody explains what data would be useful for this discussion.)
@quid I have tried to create at least some rather simple queries - just listing the questions (they are linked above). Do you think that some SEDE queries could be useful in getting some data for the discussion you were having.
I did not ping Matt Samuel since: 1. I don't see him pingable in any chatroom. 2. I created only very basic queries - probably something more nuanced should be created to get something useful.
 
7:10 AM
I have previously posted some SEDE queries related to deleted posts in my answers here: Are there some tools to follow deletions on this site? and Is there any way to see my deleted questions or answers?
However, this discussion seems to be mostly about deletion of questions that already have been on the site for some time.
What would be some useful data to quantify how often are old questions deleted? I have posted some queries above, but we can also try to create some other queries.
 
 
2 hours later…
9:15 AM
@MartinSleziak thanks. It could be useful, some care is needed though. For example, self-deletions should likely be treated separately.
 
@quid I am not really sure whether based on data from SEDE it is possible to distinguish self-deleted questions from questions deleted by roomba or by votes.
 
About half of the questions with a good score >= 5 seem to be self-deletes or other special cases.
 
(Of course, it's possible to go through them one by one; but I meant some more efficient way to find out.)
 
That's from a separate search I just performed. Namely half of them are deleted and not closed, which is a strong indicator that it was not community-deleted. (It's not a certainty but still a good indicator.)
@MartinSleziak one can do it indirectly, via score.
It'd leave out the low scoring community deleted stuff though.
 
I don't follow. How does the score tell me whether a question was self-deleted or not?
 
9:22 AM
@MartinSleziak score was regarding auto-delete
 
I see. Yes, positively scored questions were not deleted by roomba.
Similarly if a question has an upvoted answer, it was not a self-delete.
 
@MartinSleziak except if the answer was self-deleted first.
 
But I'm not sure whether from data in the PostsWithDeleted table we could be able to say much more than this.
 
I mention this since in that category of posts it does not seem all that rare.
 
@quid This is about all deleted questions or about questions deleted long time after being posted?
 
9:25 AM
@MartinSleziak all
 
AnswerCount is not in the PostWithDeleted table. So to check number of answer we'll need to use ParentId and Join two tables.
 
@MartinSleziak this only shows those that have an answer?
 
@quid Yes, you're right about that.
I wonder whether there is some workaround.
 
9:40 AM
Reducing the number of days to 400 only doubles the number.
 
hello
 
But at least from comparison the two queries we can see that most of old deleted posts with score>=5 have answers. (82 results with answers, 102 results in total)
 
@MartinSleziak with "old" in the sense of deleted long after they were asked.
 
I'll probably leave queries as they are - separate for questions with answer (which show answercount) and separate for all questions. Making a query which shows number of answers for deleted posts including 0 answers seems to be too complicated for my SQL skills.
I know that it's possible to make union of two SQL queries - but I am not sure whether the columns are allowed to be different.
@quid Yes, that's what I meant.
 
Reducing "old" does not change old that much. Even making it as low as 100 days there is not much increase.
 
9:47 AM
Which query are you testing with different "age limits"?
 
The q score 5 with answer.
 
The queries with lists are set up to shown the first 250 results - that can be changed using the num parameter.
 
@MartinSleziak yes, I did change it.
Searching for q score 5 with answer and a delay of 100 days gives 255
 
Or the one with the graphs (distribution by number of answers)?
 
The former.
 
9:52 AM
I started with three years - it seems to me that the increase effort in pushing for context is at least three years old. (And adding the missing context to the close reasons is probably even older.)
 
@MartinSleziak yes, definitely.
 
If we do not restrict for questions with at least one answer we get 243 questions for 400 days, 393 questions for 100 days. (Still keepings score >= 5.) data.stackexchange.com/math/revision/1059304/1308255/… data.stackexchange.com/math/revision/1059304/1308255/…
 
@MartinSleziak I'd speculate a lot of the difference being self-deletes.
It'll be interesting to see the numbers for answers.
 
I'll try to get list of "old" deleted questions with no answers.
@quid Simply changing PostTypeId=1 to PostTypeId=2 searches for answers rather than for questions.
 
Less than 500.
Of course posts like this one tend to inflate that count math.stackexchange.com/questions/98885/c
 
10:00 AM
I have to go AFK for a few minutes, but I should be back soon.
 
Same here actually. But it'll take more than a few minutes. Say two hours.
 
10:15 AM
No problem. I guess we have created at least a few queries which are reasonable to look at if somebody wants some stats about deletion of old questions. (I use old in the sense "old at the time of deletion" - which is why I looked at the date difference in those queries.)
Anyway, if I look at some of those questions (and also some posted in the CRUDE chatroom), quite often I would not vote to close/delete - or at least some of them seem borderline to me. So probably I have a bit lower standards than many users around here.
Of course, the questions from those queries might not be typical representative of deleted questions - considering that we are looking at questions with high enough score.
 

« first day (1640 days earlier)      last day (1778 days later) »