As a way to improve the chance of the best answers showing up closest to the question, what if you randomly reordered all low-scoring answers posted within ten minutes of each other?

Once a question (or answer) received enough upvotes you could then switch back to ordering by upvotes.

This would help reduce the impact of variations in network access/latency while still acknowledging that, all other things being equal, it's important to be able to answer the question relatively quickly.

A variation: randomly reorder all answers posted within ten minutes of the question being posted — this would probably be easier to implement....


This is a terrible way of trying to correct the FGITW "Problem" (which isn't a problem). We should be rewarding those who post quality answers quickly. If I get first in the drive-thru, I don't want to be shuffled around with 5 other customers just to make things "fake-fair".

  • By grouping answers submitted around the same time you do still reward fast replies. Random reordering within a group would give better answers a chance to be upvoted. Might improve chances of the best answer being first. Kind of like dithering. – Shea Jun 30 '09 at 20:43
  • If the first answer is not good, then it shouldn't be rewarded. But if it IS first and it IS a good answer, you shouldn't punish it for it's speed by randomly shuffling it around, which would allow a potentially worse answer up front just as much. – TheTXI Jun 30 '09 at 20:44
  • 1
    @Arnshea: That is not a good idea. The first answer /does/ deserve priority. That is why I say you don't seem to understand the issue of FGITW. – GEOCHET Jun 30 '09 at 20:45
  • Agreed, by being within the first 10minutes you'd be guaranteed to show up in the top few slots. – Shea Jun 30 '09 at 20:50
  • @Arnshea: As well you should. A fast answer is what we want on SO. – GEOCHET Jun 30 '09 at 21:14
  • @Rich, right but we don't want speed to be the only thing that matters, no? A 3 word answer in academia-speak takes less time to post but is less useful for the person trying to solve a problem who lands there from a google search. – Shea Jun 30 '09 at 21:25
  • @Arnshea: You are assuming that all voting happens in a short period of time. What actually should (and does) happen is the first pass gives the fastest and most accurate answers the most votes. Over time, voters even this out and the more detailed and helpful answers rise to the top. This is also usually accomplished via accepting the answer. You have far too simplistic a view of this process. – GEOCHET Jun 30 '09 at 21:29
  • @Rich, I'd bet that the tail is pretty short for most questions. That is, they don't get voted on again at all. And wouldn't it be better to have the first pass, which happens because the question itself is "above the fold", produce better results since, for most questions, the first pass will be the only pass? – Shea Jun 30 '09 at 21:49
  • @Arnshea: You keep implying that the first pass isn't providing good results, but what we are trying to tell you is that you are wrong. The first pass is providing the correct results, and the system is working as designed. – GEOCHET Jun 30 '09 at 21:51
  • @Rich I disagree, I think that better answers are getting short shrift because A) they take time too post and B) most people don't read past the first few. – Shea Jun 30 '09 at 21:55
  • @Arnshea: And I am telling you that this is not the case in reality, only in your head. Show me an example. – GEOCHET Jun 30 '09 at 21:59
  • @Rich The example I linked to in the above comments. – Shea Jun 30 '09 at 22:00
  • @Arnshea: There is no example or link in the comments above. – GEOCHET Jun 30 '09 at 22:01
  • @Rich, sorry, the link was in the comments to the following answer. stackoverflow.com/questions/786726/… – Shea Jun 30 '09 at 22:02
  • facepalm @Arnshea: Try to understand this. Votes have nothing to do with accepted answers. You are arguing based on nothing at all. – GEOCHET Jun 30 '09 at 22:03

I actually scan through answers before selecting one to up-vote.

  • I would +1 this if I could. Just because it is first doesn't mean it deserves an upvote, and there is nothing stopping someone from taking away a vote (except a time frame) and giving it to someone else if they choose. – TheTXI Jun 30 '09 at 20:47
  • If all the answers submitted in a 10 minute period were grouped together you might actually encourage better answers since posters would be a little less concerned about rapidly pressing the submit button. – Shea Jun 30 '09 at 20:48
  • Posters don't need to be concerned with time-restrictions....I'll seek out the best answer to upvote. – Sampson Jun 30 '09 at 20:49
  • Isn't that a little like saying there's no benefit to response timeliness? I think pretty much everyone agrees that all other things being equal the faster reply should be closer to the answer. – Shea Jun 30 '09 at 20:52
  • 1
    @Arnshea: You are contradicting yourself so much and making so little sense that you are making my head hurt. – GEOCHET Jun 30 '09 at 20:55
  • 1
    Yeah, I'm not sure where the hell you stand on the issue anymore. – TheTXI Jun 30 '09 at 20:57
  • Sorry, what I meant was, if poster's don't need to be concerned with time restrictions then that's basically saying that speed doesn't count. I think speed does count; all other things being equal between 2 answers the first should show up first. – Shea Jun 30 '09 at 21:01
  • Arnshea, I don't mean any disrespect. I just wanted to point out that I represent a demographic on SO that up-votes only answers that I think are acceptable/right. Because of this, I seek through all of the posted answers prior to picking one. My demographic isn't worried about great answers getting lumped in with a bunch of bad ones, because we hunt down the good answers. – Sampson Jun 30 '09 at 21:03
  • @Arnshea: And all we are saying is that is contradictory to your argument so far. – GEOCHET Jun 30 '09 at 21:03
  • Arnshea, speed does count, speed isn't an issue for people like me who hunt down the right answer, regardless how quickly it was submitted. – Sampson Jun 30 '09 at 21:03
  • @Jonathan, agreed but a lot of people that use stack overflow get there via google and stop reading when they hit the first answer. I'm just wondering if there's a way we can improve the chances that the first answer really is the best answer. – Shea Jun 30 '09 at 21:06
  • @Rich, I agree that speed is important, that's the reason for grouping all answers within a 10min interval. But, I'd wager that not everyone is as conscientious as Jonathan. This might cause answers to get upvoted when better answers are available. – Shea Jun 30 '09 at 21:09
  • @Arnshea: But that is the problem. A faster answer /is/ better. – GEOCHET Jun 30 '09 at 21:11
  • @Rich, what if you have 2 answers, 1 submitted 1 second after the first. The first is a one-line proof with symbols. The second is in plain english, has a short code sample and a link to more info. The first answer is better? For who? – Shea Jun 30 '09 at 21:15
  • @Arnshea, people who only read the first answer won't necessarily up-vote the first answer. If the answer is no good, they will down-vote or leave it. In the event they leave it, guys like me will come along and consider it for an up-vote among the others. – Sampson Jun 30 '09 at 21:17

Let me just say "good lord, no."

I don't think you understand the FGITW 'problem'.

  • Would you agree that, at least in principle, a system that put the best answers closest to the question would be preferrable to one that didn't? Unfortunately some aspect of being close to the question is pretty variable (network location, filtering, proxying, etc...). – Shea Jun 30 '09 at 20:46
  • Arnshea: If you sort your answers by number of votes you'd realize that's how the system already works... – TheTXI Jun 30 '09 at 20:47
  • @Arnshea: The way it works now accomplishes that. Your way would only go away from that, not closer. – GEOCHET Jun 30 '09 at 20:48
  • How do you account for situations like stackoverflow.com/questions/786726/… ? The second answer has much more detail and should probably be the first answer. But it took longer to post so didn't get the "above-the-fold" treatment. Grouping might have changed the outcome for the better. – Shea Jun 30 '09 at 20:58
  • @Arnshea: They have the same number of votes. What am I missing? Do you not understand the accepted answer mechanism? – GEOCHET Jun 30 '09 at 21:00
  • @Rich, I've seen other people who get to stack overflow from a google search. They don't even know they're at stack overflow, they just look at the first answer (the one in green); if it helps them great, otherwise they click "Back" and look for other answers. – Shea Jun 30 '09 at 21:27
  • @Arnshea: Ok. And how does that change anything? Do you understand what an accepted answer is? Hint: It has nothing at all to do with votes. – GEOCHET Jun 30 '09 at 21:30
  • @Rich even the person who wrote the question might be tempted to only read through the first few answers to decide which they'll accept. I think we can improve the quality of the first few answers with randomization. The same way you can improve the accuracy of naval guns with dithering. – Shea Jun 30 '09 at 21:51
  • @Arnshea: Provide the evidence. Where are all these questions with crappy answers on top and better answers on the bottom? – GEOCHET Jun 30 '09 at 21:55
  • @Rich, I linked to one example case, there are many others but I'm not motivated enough to link to them all. – Shea Jun 30 '09 at 21:56
  • @Arnshea: The 'example case' you gave didn't match your argument at all. The answer on top was the accepted answer. What do you not understand about accepted answers? – GEOCHET Jun 30 '09 at 21:58
  • @Rich The accepted answer was not the best answer! – Shea Jun 30 '09 at 22:01
  • @Arnshea: Oh lord. You really don't understand accepted answers. Accepted answers have nothing to do with voting. They are chosen by the OP. Your argument is based on your lack of understanding of a simple mechanism on the site. Please read the FAQ. – GEOCHET Jun 30 '09 at 22:02
  • @Rich, Yes but the better answer would probably have gotten more votes if it appeared earlier. This would happen, on average, less often if the answers submitted within a time interval were randomly reordered. – Shea Jun 30 '09 at 22:05
  • @Arnshea: The 'better answer' /does/ have more votes. This still has nothing to do with the accepted answer. I have seen answers with negative votes accepted. Why don't you understand this disconnect? – GEOCHET Jun 30 '09 at 22:08

I actually like this idea ... group responses and randomize (or sort by votes) within certain time frames (perhaps 1st 10 min, 1st hour, 1st 8 hrs).

A clear and public goal of Jeff and Joel is for each of these sites to be the "canonical" home for answers in it's focus area. To accomplish that, the site should encourage thorough, careful, well-thought out and well written responses. A corollary of the "canonical site" goal is that questions and answers need to have applicability beyond the immediate need.

Quick, shoot-from-the-hip answers get votes, but the thorough, complete answers address the long term goals of the site. The FGITW pattern is a problem because quick, incomplete answers are rewarded with votes and rep, effectively discouraging the kinds of answers which the site really needs.

The proposal mitigates the issue by reducing the advantage of being first while still giving some advantage to being among the first.

  • FGITW is not a problem. Show us all these good answers on the bottom and bad answers on the top. – GEOCHET Jun 30 '09 at 21:56
  • THANK YOU!! Excellent description of the problem! – Shea Jun 30 '09 at 21:57
  • Yeah I think some people are forgetting that the goal of the site isn't rep. Rep is just a carrot. It's how the site's creators get us to do the work without paying us in money. – Shea Jun 30 '09 at 21:59
  • 1
    facepalm Where is that badge? – GEOCHET Jun 30 '09 at 22:00
  • The thing is, because you would slightly disincent the quick "shoot from the hip" answers, the overall quality of answers would improve. And Rep would still be just as motivational. – Shea Jun 30 '09 at 22:06
  • @Arnshea: You need to lurk more then. Most quick answers are fleshed out later to be more comprehensive, so your argument has very, very little merit (if it has any at all). – GEOCHET Jun 30 '09 at 22:10
  • @Rich, I bet you're wrong on that. I bet most questions get very few views and far fewer votes. So given that most questions only get a few minutes/seconds of fame, randomization might improve the correlation between #votes and answer quality. – Shea Jun 30 '09 at 22:14
  • @Arnshea: Obviously you are not familiar with how the system bumps random questions to the front page among the other ways questions get additional attention then. – GEOCHET Jun 30 '09 at 22:16
  • @Rich, right and how much do you think that ups the votes on a site with over 200,000 questions? – Shea Jun 30 '09 at 22:21
  • @Arnshea: Quite a bit. But if the problem is as pronounced as you seem to think it is, then surely you can provide one example, right? – GEOCHET Jun 30 '09 at 22:23
  • @Rich, there are only 86400 seconds in a day. That's less than half a second of "fame" for every question. Then add in the fact that question answerers and voters probably account for less than 1/3rd of the total hits. There just isn't enough time in the day for occasionally bumping a question above the fold to do the trick. – Shea Jun 30 '09 at 22:28
  • @Arnshea: Where is that example again? – GEOCHET Jun 30 '09 at 22:30

You must log in to answer this question.

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