Not sure if this is being done already, but...

There seems to be a problem with the "hot questions" list, in that it almost always has a lot of questions from the Math.SE site. I imagine two possibilities:

  1. All SE users are really into their Cauchy sequences and algebreic theory
  2. The Math.SE users are quicker to upvote than their sister sites, and the number of upvotes somehow biases the system to thinking that the question is popular across sites

I've seen this with Academia as well (I'm a mod there); we tend to upvote a lot, and during the semester, we'll have a fair number of questions in the Hot Questions site just because we tend to upvote a lot. This is actually somewhat borne out in the data; The score distribution on the main SO site is very different than that on Math.SE, for example (x-axis is score, y-axis is post count; it's a crude plot without many exclusions, feel free to fork and improve).

To that extent, I would recommend a scaling system. We're nerdy here; we can do this appropriately.

  1. Plot the score histogram for each sub-site
  2. Assign one of the histograms as the "reference" distribution, i.e. the one that all other sites should be scaled towards.
  3. For each question score, determine how much that question needs to be scaled such that it resembles the reference distribution.

In other words, if we choose the two plots above and call the SE one the "reference" one, we could scale the votes in question using the formula shown in column F below:

enter image description here

Note that we could do the same exact thing for "speed that upvotes are given"; i.e., on Math, questions accumulate upvotes much more quickly than on other sites, and we can normalize that distribution as well.

  • 1
    Surprise, surprise. A chart from someone at math.se.
    – Kermit
    May 10 '13 at 20:09
  • 3
    @FreshPrinceOfSO I'm from Academia, thankyouverymuch. I don't understand jack on the Math SE.
    – eykanal
    May 10 '13 at 20:09
  • 2
    *A chart from someone that does mathy things.
    – Kermit
    May 10 '13 at 20:10
  • +1 for 'We are nerdy here we can do this right'. But really +1 because I was thinking of writing this up the other day. It's also an issue with the tex.se sight. I absolutely can't believe that LaTex is as popular as the hot-list makes it out to be.
    – Ryan
    May 10 '13 at 21:07
  • 4
    Have folks not realized yet that the Hot Questions list is worthless due to the bizarre definition for "Hot"?
    – user7116
    May 10 '13 at 21:12
  • Let's fix this by inspiring SO users to vote like Math users. May 11 '13 at 0:42
  • @sixlettervariables - For being worthless, it sure is given a very prominent position in the site UI.
    – eykanal
    May 13 '13 at 13:34
  • would be interesting to estimate how Math questions would rank at collider if formula would discard answers when voting evidence indicates that these are not good data points -- ie those "having score less than (TopAnswerScore/10-1)". Per my observations, Math guys indeed upvote a lot, but they don't tend to upvote low quality answers (why would they?)
    – gnat
    Nov 8 '13 at 13:13

You must log in to answer this question.

Browse other questions tagged .