Timeline for Reorder questions picked for hot list based on adjusted hotness score (discard some answers by voting evidence)
Current License: CC BY-SA 3.0
19 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 3, 2020 at 13:30 | history | edited | CommunityBot |
Commonmark migration
|
|
| May 23, 2017 at 12:36 | history | edited | CommunityBot |
replaced http://stackoverflow.com/ with https://stackoverflow.com/
|
|
| Aug 24, 2014 at 14:02 | comment | added | gnat | ...however I am also free to decide whether to accept an explanation for decline or not | |
| Aug 24, 2014 at 14:01 | comment | added | gnat | you make some interesting points that I didn't consider these when making this feature request, I even initially accepted because of that. However I recently learned that SE doesn't even keep simple log of what happens in the hot list. This makes it look like pure speculation doesn't it. I mean we can't tell if your reasoning is right or wrong because we simply have no data to help us decide. That's... sad. You are of course totally free to pick the reason to decline... | |
| Jul 6, 2014 at 11:57 | vote | accept | gnat | ||
| Aug 24, 2014 at 14:02 | |||||
| Jul 6, 2014 at 11:57 | comment | added | gnat | correct if I misunderstood, the primary reason for decline is that suggested feature doesn't address the difference of thought provoking questions from bikeshed and trivial ones. (Gee it took me almost half to even discover this point. Can't tell if this is because it was obscured by other considerations or maybe I was simply blinded by the passion to tweak hotness formula) | |
| Feb 14, 2014 at 16:45 | comment | added | user194162 | Thanks for the additional breakdown of hot questions and the related effects that could potentially occur from changing the algorithms. | |
| Feb 13, 2014 at 5:47 | comment | added | gnat |
as far as I can tell performance cost of suggested feature is acceptable, is it? Also you mentioned that simulations show that impact is small and so are associated risks, correct? (btw minor impact is intentional; for a bit more impact/risk you can consider cut at something like TopAnswerScore/5-2 instead) Wonder why you don't just give it a try, are you afraid of something? Are you afraid that it will succeed and break your dream of completely dropping hot list? Or that it'll fail and I will push for hot list removal myself?
|
|
| Feb 11, 2014 at 17:57 | comment | added | Shog9 StaffMod | If you refer to the formula that drives this, you'll see that the only way a question can end up sticky is to maintain a score higher than most of the other questions on the network - either by accumulating answers and/or votes at an unusually high rate, or by simply not having any real competition for a good long while. Think of it as one balloon floating higher than the rest of a bunch. So, you tie a bit of lead to the string... @gnat | |
| Feb 10, 2014 at 7:57 | comment | added | gnat | I don't understand how penalty is supposed to help when particular question sticks to top and starts quickly collecting 5...10...20 lemming answers? Also, what about other sites complaining about similar issues (Programmers, Math, Code Golf, UX...)? are these also to be slapped | |
| Feb 10, 2014 at 7:23 | comment | added | Shog9 StaffMod | Yeah, that's... Unlikely to do much of anything at all. We could slap a fixed penalty on questions from TWP and have a much greater effect on what shows up in the list with considerably less effort and overhead. | |
| Feb 10, 2014 at 5:54 | comment | added | gnat | yes, this request is limited only to 100 questions that have been "pre-selected" using current formula. ("Global" score adjustment has been suggested, discussed and declined by David per prior request.) The point is to manage sidebar exposure: take 100 questions already selected as of now, adjust their score, reorder and feed top 3-to-26 into sidebar. | |
| Feb 9, 2014 at 22:10 | comment | added | Shog9 StaffMod | No. That's the perf cost if we don't actually expect the list to change based on the altered criteria. Normally, those 100 results are pulled from a much, much larger pool of questions, with the "hotness" score calculated for each of them so that they can be ranked accordingly. But if we're expecting the same 100 results and just want them ordered differently, then this would suffice - in fact, you could probably estimate the cost of this using the public API. Of course, if we're just getting the same results, then... what's the point? | |
| Feb 9, 2014 at 22:05 | comment | added | gnat | I see, all right. Let's see if I understand it correctly this time: performance cost of this feature request is that of re-querying 100 questions (we're talking these that are listed) with answers ordered by votes - is that right? | |
| Feb 9, 2014 at 21:46 | comment | added | Shog9 StaffMod | "The same data" doesn't include individual answer scores or close votes. You have the site, creation date, score, last activity date, answer count, deletion date, closed date, tags and answer score. That's it. | |
| Feb 9, 2014 at 21:45 | comment | added | gnat | you query some data already don't you - to create these funny looking tooltips in hot network list: "This question has been arbitrarily awarded 100.982 hotness points". Feature request only suggest to use the same data, that is already queried and used, nothing else | |
| Feb 9, 2014 at 21:38 | comment | added | Shog9 StaffMod | You're kinda missing the point here, @gnat: it's not that this data is slower to obtain (although it is) - we'd have to query every site's database in order to even get it. That's just not feasible - implementing this would mean denormalizing additional data and somehow keeping it updated quickly enough for it to be meaningful, which seems like a huge amount of work for something that would almost certainly have detrimental effects on the results anyway. | |
| Feb 9, 2014 at 21:35 | comment | added | gnat | nothing challenging performance wise is needed over what is there now, because "hotness correction" is limited to posts already picked by current algorithm. Take 100 questions and less than 1000 answers already selected as of now, adjust their score, reorder and feed top 3-to-26 into sidebar. That's it. O(1), performance "cost" is negligible | |
| Feb 9, 2014 at 21:32 | history | answered | Shog9StaffMod | CC BY-SA 3.0 |