Possible Duplicate:
Count only questions with upvoted answers towards the accept rate percentage

Possible Duplicate: Count only questions with upvoted answers towards the accept rate percentage

The accept rate is computed based on questions that have answers and serves (as I understand it) to judge how likely it is for a user to care about accepting answers.

To give that rate more credibility, I propose basing it on the likelihood that a user will accept good answers. It's not entirely obvious how to quantify "good", but one way could be specifying a minimal score like 0 or rather 2 (which is the bounty auto-award threshold).

So, to determine the accept rate, only questions with at least one answer with score k (e.g. k==2 or k==0) or higher should be considered.

I think that would remedy the valid criticism of the accept rate that comes up from time to time.

  • 2
    Meh, I don't think you can possibly attempt to quantify a "good" answer across all of the tags on StackOverflow. Just because two other people didn't up-vote it doesn't mean it's not a good answer. I think the answer is to just stop making the accept rate so prominent, since it's largely irrelevant - accepting answers is good, but not compulsory. Like "good" or "poor" answers, you can't determine at a glance that an accept rate is low because they asked terrible questions, didn't bother accepting good answers, or didn't receive satisfactory answers. – Aaron Bertrand Aug 21 '12 at 16:56
  • 9
    We should just get rid of the damn thing, it's as annoying as flag weight used to be... – yannis Aug 21 '12 at 16:56
  • 6
    @Yannis - maybe. But don't you think a 0% accept rate is often symptomatic of a help vampire? I think the problem is that virtually any comments about accept rate get destroyed, which leads to a lot of time-wasting meta posts. – Adam Rackis Aug 21 '12 at 16:57
  • 1
    ...indeed, annoying. But I wonder, @Yannis, might folks who hit the daily cap look at it? It's their only way to earn some reputation, for those who're into that. – Arjan Aug 21 '12 at 16:58
  • 1
    The proposal is sound, but I disagree the score limit of 2. In "niche" tags like JSF/JSF2, answers rarely get a score of more than 1. – BalusC Aug 21 '12 at 16:58
  • 1
    @AdamRackis i.stack.imgur.com/rh6su.png ;) – yannis Aug 21 '12 at 17:00
  • @Chichiray: It was just an alternative to 0. You can come up with arbitrarily sophisticated means to determine "good" -- you could take the average score on the question's tags into account and so forth. But personally, I prefer simpler rule sets. – bitmask Aug 21 '12 at 17:01
  • @Yannis - well played :) And Jon Skeet has a low accept rate too, but in general... – Adam Rackis Aug 21 '12 at 17:01
  • @AdamRackis A low accept rate may be a good hint of a help vampire, but unfortunately I think there are just too many users obsessed with getting a few more imaginary points pestering newer users about it. I feel that showing the accept rate has failed its purpose and we should either replace it with another, less visible perhaps, mechanism to detect help vampires or if get rid of it completely. – yannis Aug 21 '12 at 17:14
  • replace it with another, less visible perhaps, mechanism to detect help vampires or if get rid of it completely - I'd be in completely in favor of that. – Adam Rackis Aug 21 '12 at 17:16
  • @YannisRizos: Perhaps the answer/question would be better suited to detect help-vampires, then. If you ask 20 questions, never answer a single one your are not giving back that much, right? – bitmask Aug 21 '12 at 17:17
  • @bitmask The quality of the questions should factor in, great questions are beneficial to the community even if you never posted an answer. – yannis Aug 21 '12 at 17:23
  • 1
    I think k should be -1, i.e. only answers with very high probability of being incorrect should be disqualified. – dasblinkenlight Aug 21 '12 at 17:24
  • @dasblinkenlight: Fair enough. I just proposed 0 and 2 as examples. – bitmask Aug 21 '12 at 17:25
  • @bitmask plenty of people are great contributors to the site even without ever answering a single question, so I think your statement that never answering a single question means you are not giving much back is incorrect. – Aaron Bertrand Aug 21 '12 at 18:28

Bad questions attract poor answers. Accept rate is, to a significant extent, a measure of question quality.

In my experience, a 0% accept rate just means that the OP doesn't know how to accept answers yet.

  • I absolutely agree with your statement about 0%. However, I've more than once accepted an answer just to keep my accept rate in the green (i.e. >80%). That feels like broken design. – bitmask Aug 21 '12 at 17:41
  • @bitmask Any accept rate above 50% is OK. Any accept rate above 70% is excellent. See meta.stackexchange.com/a/16729 – user102937 Aug 21 '12 at 17:44
  • Fair enough. – bitmask Aug 21 '12 at 17:55
  • If you're accepting just for the accept rate, @bitmask, then see "What should I do if none of the answers is suitable?" in How does accepting an answer work? – Arjan Aug 21 '12 at 18:34
  • @Arjan: I do this sometimes. By no means all the time. Also, mostly not on SO. – bitmask Aug 21 '12 at 18:59

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