27

I've recently noticed that the hot questions list is now different almost every time I look at it. It used to be static - I.e. I could browse around for half an hour and the items on it would be minimally changed.

Was the algorithm changed? In what way?

  • 1
    Really? I honestly find they stick around for longer. There was at least one question that hung around for almost two weeks on my end. – Eric Feb 1 '14 at 0:30
  • 1
    This is actually a very recent change, like within 24 hrs of now if I'm correct, @Eric. – Undo - Reinstate Monica Feb 1 '14 at 0:32
  • 2
    Fair enough, I haven't been online enough in the last 24h to know for sure. (I do hope they've changed it as it was seriously favoring sites like Code Golf; though, considering their decline in V/D, I can see that the algorithm has probably changed.) – Eric Feb 1 '14 at 0:34
  • dont think thats a desirable design feature that it seems somewhat random on every refresh.... apparently there are many hot questions to choose from, but how could that be? doesnt that sort of negate the unique status of hot questions? – vzn Apr 2 '14 at 15:48
  • 1
    @vzn 200 hottest questions, select 50 at random from that list. Out of tens of thousands of questions that still pretty hot but you don't see the same questions over and over (which is boring) – Richard Tingle Apr 2 '14 at 15:59
42
+50

The usual algorithm calculates a score, ranks the questions and then just pulls off the top few results.

We're testing an alternative method that calculates a score, ranks the questions and then pulls a somewhat larger number off the top, shuffles them, and stuff the top of the deck into the sidebar.

Since this is a test, not everyone will see this. Those who do will observe both a larger variety of questions, a less predictable order for those that do show up repeatedly and a greater chance of seeing different questions on different page-loads.

And since this is a test, it may end up having horrible side-effects that result in it being tied up in a sack with a brick, thrown off a bridge and never spoken of again.

  • 1
    A/B testing? (I don't see it, currently) – Doorknob Feb 1 '14 at 0:39
  • 8
    Yeah. If the list doesn't change every time you refresh the page, you're on the B team. Or maybe it's the A-Team. I forget whether or not Hannibal played cards. – Shog9 Feb 1 '14 at 0:40
  • 1
    If it doesn't work, will the employee(s) responsible also be put in said sack? – Eric Feb 1 '14 at 0:48
  • 18
    They'll be given a llama and told to pair-program with it. – Shog9 Feb 1 '14 at 0:51
  • 1
  • 3
    I love this change oh-so-much. This is a much better way of getting exposure to other sites and definitely decreases the likelihood of a 40-year flood every few months compared to the previous system. Plus it shows you guys care about this issue, which is definitely awesome. Thanks Shog! – jmac Feb 3 '14 at 1:37
  • 1
    Thank you Jeremy and @Jarrod! – jmac Feb 3 '14 at 1:39
  • 2
    I shoulda just credited Jon Skeet, since apparently it's his implementation of the shuffle algorithm at work here. And also, because there aren't enough J-names involved in this thread yet. – Shog9 Feb 3 '14 at 2:15
  • 1
    great experiment! To help in testing, I plan to visit Stack Overflow questions in the hot list and actively vote up posts I like. This way will let SE team study how effective it works against "lemming effect" using familiar material (instead of obscure questions at smaller sites like Programmers, Workplace, Math, Code Golf, UX) – gnat Feb 4 '14 at 5:48
  • 2
    This seems to have been shut off. I was rotating questions the past few days, but it seems the latest build(?) may have shut it off again for great sadness. Can we get a follow-up on why and future plans please? – jmac Feb 6 '14 at 7:20
  • 1
    @Shog9 Disappointing to hear it's not working out so well. I enjoyed quick access to a whole bunch of interesting questions from various sites this week. Now that I appear to be back in the control group, the list has returned to 1-2 good questions and a whole bunch of "meh" clogging up the slots for days on end. – Troyen Feb 7 '14 at 23:44
  • 1
    just a thought about significantly smaller number of people are clicking through... maybe it's time to to decide whether number on clicks is so much more important than the kind of promoted content. After all, if SE team is so much into the racket of clicks, most natural direction would be to shift to porn/sex and dump the idea of attracting experts in sites topics; it's hardly a coincidence that recent sticky top-hot question had sex in the title, otherwise being pretty mediocre – gnat Feb 8 '14 at 19:45
  • 1
    I am merely pointing that it's absurd to use number of clicks as the only criteria in the test. It is also absurd, up to the point of becoming totally kafkian to run that test without even getting an understanding of what is the purpose of hot questions. Given that, I am not surprised to see that you consider answers-up as a good sign... – gnat Feb 8 '14 at 20:43
  • 1
    @Shog9 I would possibly buy that if only I could see; if I could see that trouble brought by genuinely hot questions correlates with site quality issues, I'd probably join your crusade. But the way how things work now doesn't allow for that - hot questions are fake way too much and way too often to make any meaningful conclusions based on them. Maybe that's why you're alone in your quest for HQ removal, people just suspect that you're throwing out the baby with the bathwater – gnat Feb 8 '14 at 21:19
  • 3
    This post is referred to quite widely across this meta. Could you add a quick update to the answer stating whether this has been definitively enabled across the board? – E.P. Dec 3 '16 at 16:48

You must log in to answer this question.

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