3

Two questions:

  1. Is there anything in the HNQ algorithm that explicitly prevents a question that was on the list from appearing again at a much later date? E.g. a hard age cutoff or anything?

  2. If not, is there an actual risk of this happening / has it ever happened in on a significant scale?

The reason I ask is I am writing a little HNQ tracking app, and the logic that I use to track "time on the list" for questions near the end of the top 100 that may pop off and on it depends on this. So if my questions above aren't the right questions in this context, my real, underlying question is:

  • If my software defines "time in list" as "last seen in list" - "first seen in list", will this always be sufficient? Or will I have to do more complex logic to avoid weird things like "6 months in the list" if a question reappears?

According to https://meta.stackexchange.com/a/61343/230261, the hotness formula (not including traffic weighting), is:

(MIN(AnswerCount, 10) * QScore) / 5 + AnswerScore
-------------------------------------------------
         MAX(QAgeInHours + 1, 6) ^ 1.4

Assuming there are no explicit age limits or other logic, this means that a question may reappear on the HNQ list if its score or answers ever get that weight back to the original x / Age^1.4 line. This seems unlikely.

6
  • It's actually not that rare. Get a question and answer combo with 3 upvotes each within six hours, and most of the time it will hit the bottom of the HNQ list. It will drop off pretty quickly after that, but if a fourth upvote or second answer comes in shortly thereafter, it will show up on the list again. – Nathaniel is protesting May 7 '17 at 3:09
  • @Nathaniel Yeah; that's OK, that's actually the reason for the question: I know that the ones near the bottom of the top 100 will pop off and on the list since they're right on the border. So the simplest way to handle that is "last seen time - first seen time", which fits my definition of "on the HNQ list". But if I handle it that way and a question disappears then comes back like two months later, then I'd falsely treat it as being on the list for 2 months. Hence this question. I want to know if they can come back much later, so that I can handle the borderline fickle ones simply. – Jason C May 7 '17 at 3:27
  • 1
    I'm pretty sure that nothing checks if a question was ever HNQ in the HNQ algorithm, it simply does not care. So the only way to be 100% sure is to scrape the list every 15 minutes, which as far as I know is its caching time, then just add 15 minutes of "Total Hotness Time" to each question you find there. – Shadow Wizard is Vaccinating May 7 '17 at 13:34
  • Turns out this is totally answered here but I missed it: "Questions older than 30 days are excluded". So no question will ever be there longer than a month, or will ever reappear more than a month later. – Jason C May 7 '17 at 14:17
  • @ShadowWizard I like your approach of counting total time instead of just ignoring gaps. That seems to make more sense. Although I'm torn: While that's technically probably more accurate, as far as us humans are concerned if a question pops off and on the list a lot we perceive it as being there the entire time (a borderline question that only had 1 day total time over 3 days would still be looked at as "man that's been there for 3 days now"). Maybe I'll track both. – Jason C May 7 '17 at 14:19
  • @Jason yeah, catching both seems ideal. "Real Time" vs. "Seen Time" or something like that. :) – Shadow Wizard is Vaccinating May 7 '17 at 14:38

Browse other questions tagged .