7

To start with, I am aware that hot questions from larger sites are somehow "penalized" and what are the reasons for that:

This is designed to ensure that a single site doesn't dominate the list, and that users see a good sampling of questions that are likely to interest them throughout the network...

I would like to understand how is hotness score really calculated at these sites. The reason why I ask is that per my observations, this seems to be quite different than is documented in the description of "hotness formula". Specifically, I would like to understand the difference in the impact of answer(s) score and aging factor.

Regarding impact of answers score, my observations of hot questions from smaller sites appear to be in agreement with known formula: (MIN(AnswerCount, 10) * QScore) / 5 + AnswerScore. For example, a question with low score (+1, +2) can get into hot list if it has got answer(s) with sufficiently high score.

At larger sites, this doesn't seem to be the case. Say, it looks like Stack Overflow questions just don't enter hot list until they reach certain score (+6,+7), no matter how many answers are there and how high is their score.

Does system simply totally ignore answers until question reaches certain score?

Another thing I noticed is that questions seem to age away from the hot list 2-3 times faster than ones from smaller sites (having comparable question and answers scores). Granted, this is harder to track and I am not yet sure if I figured that correctly. Do larger sites questions age away faster than those from smaller sites?


For those interested, here is a (partial) list of example questions I studied before asking this: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40.

Although above examples are only from Stack Overflow, my question isn't (I only used it as an example of a large site in my studies). I would like to understand if hot questions from other larger sites are handled in a similar way.

2
12

Rough pseudocode:

IF siteId in (over-represented sites) OR voters include (gnat) THEN Score = Score*0.2

May not be entirely accurate, but close enough to give you a rough idea of how primitive this is.

There's also an additional penalty applied to questions if a previous question from the same site is already in the list. This tends to dramatically reduce the chances of multiple questions from the same site appearing, even if multiple questions have the same stats at a given point in time, and is usually sufficient except in cases where someone's running around upvoting all the terrible answers to crap questions.

8
  • that additional penalty, it is documented as "the first question from SO in the list gets multiplied by 1.0, the second by 0.98, the third by 0.96, etc". I assume this is up-to-date, is it? FWIW I didn't observe it making much impact, neither at small nor at large sites
    – gnat
    Jan 11 '15 at 18:49
  • The idea hasn't changed, I'm not sure those exact multipliers are correct though.
    – Shog9
    Jan 11 '15 at 19:11
  • 7
    Careful, mods will be asking gnat to vote on questions we don't want in Hot Questions now... :]
    – user206222
    Jan 11 '15 at 19:39
  • thanks. Approach like you describe seems to match my observations pretty well. (Side note if impact of votes from a single user feels worrisome, maybe it's time to think of a more reliable feature implementation)
    – gnat
    Jan 11 '15 at 20:43
  • 1
    Mostly joking there, inspired by a fairly ridiculous method I've encountered of late for detecting voting fraud on SO (abysmally bad question in hot list? Probably a ring of sock/meat puppets behind it).
    – Shog9
    Jan 11 '15 at 21:01
  • ah, Bitcoin.SE type of fraud. I see
    – gnat
    Jan 11 '15 at 21:24
  • FWIW statement that "additional penalty... tends to dramatically reduce the chances of multiple questions from the same site" is quite outdated. Now that all questions make a randomised list at sidebar and there is hard limit of 5 from same site, this penalty "fails to make any reasonable impact". Casual sampling of hotness scores in some list snapshot makes it clear that even maximum currently possible penalty of 8% merely shifts the question 10-15 places down the list (of 100) which hardly qualifies as dramatic reduction of chances...
    – gnat
    Jul 30 at 12:23
  • ...this parameter value was intended to work (and possibly worked well enough) in ancient times of "multi-collider dropdown". System has changed too much for it to make sense now
    – gnat
    Jul 30 at 12:23

You must log in to answer this question.

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