14

Most people have a bad day once in a while and post an answer that doesn't receive any upvotes. With an increasing number of posted answers the probability increases that there are such answers. If now somebody posts several answers and receives an upvote on every last one of them, this might be an indication of gaming.

Of course some people legitimately get upvotes on nearly every one of their answers, but usually then the average score of these answers will also be rather high. For example Joel has an average answer score of 11, which can't even be explained by the whole SO staff serially upvoting him.

But if a very high rate of upvoted posts comes together with a rather low average score of these answers, it looks suspicious and might warrant some moderator to pull out some advanced database queries.

For example, if you pick this guy, who got at least one upvote on every one of his answers but only an average answer score of 1.25, you find that his latest >10 answers were all to questions by a single account. Most of these answers were also accepted. This looks like it would be worth further investigation by a moderator.

For more of these case here is a data explorer query looking for users with high upvote rates on their answers, but low average score (ordered by some rather arbitrary metric). So if there are some internal database queries that analyze voting patterns, many of these users might be good candidates for further investigation.

The query might also give quite improved results if it would include deleted posts, since users might delete answers that are not receiving any votes, but the data explorer doesn't contain that data.

1
  • "this guy" was just merged with another guy :) +1 from me, always looking for interesting queries on data explorer
    – waffles
    Commented Jul 14, 2010 at 6:14

2 Answers 2

3

Some 'low-activity' tags, with only a handful of people answering the questions, might legitimately get a lot of correct answers with one or two upvotes.

2

I'm not a statistician, but maybe you could look for people where their votes don't obey a Poisson distribution because they have too few highly upvoted answers.

1
  • +1 I had the same thought, although it'd make sense to start by trying to establish the usual distribution of votes (i.e. don't just assume it's Poisson - I wouldn't be surprised if it is, though)
    – David Z
    Commented Jul 14, 2010 at 7:30

You must log in to answer this question.

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