What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 133 Stack Exchange communities.

What are thoughts on adding the following two statistics to the user profiles, possibly only to be displayed for the logged in user (i.e. other users can't see these stats):

  • Average answer value - this would be the average total score in reputation points for all answers the user has left. Reputation gains for accepted answers should be counted here as presumably an accepted answer is thought to have higher value than one that isn't.
  • Answer acceptance rate - this is the percent of answers posted by the user that are ultimately accepted. It should only count questions that are eligible to have an accepted answer. Which probably won't be a problem if these statistics are cached anyway.

Personally I would find these values interesting and they would hopefully serve to motivate a user to improve their average score by being more conscious of writing high-quality answers.

As a rough example using my profile which is only taking my total reputation / questions answered (realizing that I've gained reputation elsewhere...) my stats would look like (accept rate from SEDE):

  • Avg. Answer Value = 26.52
  • Accepted Answer Rate = 31.36%

EDIT

In light of Anthony's concern, perhaps the average answer value should be average answer score, which would be the average sum of upvotes and downvotes instead of reputation score. In this case upvote and downvote have equal weight where in the reputation calculations they do not (10 vs 2).

share|improve this question
    
On the other hand, it might also dissuade prolific users from providing too many helpful answers, as the rep cap effectively drops their average total reputation score. Jon Skeet would look like a mere mortal, which is purely nonsensical. –  Anthony Pegram Jan 12 '12 at 16:03
    
@AnthonyPegram Kind of why I feel they should be kept private. Also why the metric should be calculated from votes and not purely actual reputation as I've done in my cheap example above. And I think Skeet's ~400k rep speaks for itself :) –  Yuck Jan 12 '12 at 16:06
5  
Average answer score and acceptance percentage are metrics I can rally behind. –  Anthony Pegram Jan 12 '12 at 16:12

4 Answers 4

I somewhat agree with your proposition, percentages of things would be nice. But I don't see much importance in average answer value and accepted answer rate. This would be nice, but I see several other percentages as a priority.

  • Vote percentages on answers: This would just be the percentage values of the +1's you've received and -1's youve received. Sort of like an acceptance rate, like: This user has 67% up votes on answers, and 33% downvotes on answers.

  • Vote percentages on questions: Exactly the same as stated above, but these two need to be seperated.

  • Average answer score: You answer alot of questions (77 for example) and throughout this process, you have an average of 5 on answer scores.

  • Average question score: Take a guess

  • Percentage of answers/questions in specific tags (this could be a pie chart or something!) So if you answer alot in c++ tags, youd have maybe 78% in c++ tags, 12% in java, etc.

  • Your recommendations as stated in your question

Just to show an example of what I mean by all this, check out the illustration below

enter image description here

share|improve this answer
1  
Very nice, glad to have helped inspire this and I appreciate the mock up to illustrate what the feature might look like. –  Yuck Jan 12 '12 at 16:27
    
@Yuck Yeah I love your question because I've always thought a little statistics tab would be great. I had to just throw my ideas out there –  Gabe Jan 12 '12 at 16:28
1  
Regarding the pie chart, careful with the tag overlaps, or it will need more than 360 degrees. (I mean, generally, it tag scores shouldn't be expected to add up to 100%.) –  Bruno Jan 12 '12 at 16:52

I'm not entirely sure what this would reflect. I would already dispute the (linear) correlation between points and "merit" to a degree.

Not all points are equal. The more specific the topic is, the harder it can be to get upvotes, even for correct, clear and documented answers.

For example, I'm not saying that all my answers are clear and perfectly documented (I'd like to think they're correct), but I've answered a lot of questions in the ssl tag. Most of them required a bit of expertise to answer. I think the fact that the subject can be quite specific (both for questions and answers) makes it difficult for non-expert to assess whether or not the answer is good or bad, and thus they wouldn't necessarily upvote answers (not knowing whether it's correct or not). There's nothing wrong with that (I tend not to upvote answers that may look correct but when I don't feel confident enough to be sure).

This results in a large number of answers with few upvotes, despite them being often accepted. The counter-part of this specificity downside is that it makes it easier to get the Unsung Hero badge.

The metric you're suggesting clearly puts an emphasis on what I'd call "easy" questions. One of my highest scoring answer to date (including points for accepted answer) is this, which, frankly, any programmer with a few days of Python training could have produced.

To improve my average score, a good strategy would be to look for easy questions and answer them as soon as they're posted. It would seem to work better than taking the time to answering more specific questions that require quotes from detailed documents.

share|improve this answer

I agree, was about to submit the same suggestion. It seems that most metrics on the site revolve around reputation. I am much more interested in whether the asker accepted my answer. I'd much rather have an accept with 0 votes than a non-accept with 10 votes. I want to know that I helped solve the original problem, not that a bunch of casual observers liked my solution. I realize the latter should be helping the OP with the former, but this is not always the case.

I would be quite happy with just totals (since I can do math). Sure I can figure out my own accept rate, if I'm really curious, by paging through all ~800 answers, 30 at a time, and tallying how many have a green box. If that sounds like fun to you I also have some yard work that needs to be done. I know the database can do this for me with very little effort at all. Right now I just take a glance at "how am I doing lately?" E.g. on my most recent 30 answers, have 10 or more been accepted? If so, I'm happy.

On MSDN Forums they show totals for "Answers" (an accepted answer in our terminology), "Helpful posts" (analogous to our upvotes), and "Replies" (which would be our answers). I think this would be useful to show on our profile as well. While I agree with @Bruno that quantity does not necessarily indicate quantity, I'm not talking about judging other users based on these numbers, I'm just talking about judging ourselves (e.g. I am fine with the earlier sentiment that we only see these numbers for ourselves - if someone else wants to calculate this for another user, they can do it the hard way).

How MSDN Forums do it

share|improve this answer

I think the average answer score would be a valuable metric. Consider two users with 1000 rep; one has posted a small number of very highly "scored" answers, while the other has posted a massive volume of near-0 score questions. One really should have a higher "reputation" than the other; as it stands it requires a (very small) amount of work to figure out which is the more reputable user.

share|improve this answer
2  
It's not that easy to say one would be more "reputable" than another. Sometimes, highly upvoted answers don't necessarily reflect their quality, but rather the fact a large number of users have had the same problem. Let's take this answer: 991 upvotes, yet anyone googling for "json mime rfc" or reading the JSON wikipedia entry would have found the answer. I'm not saying the answerer doesn't deserve a high reputation, but does this answer (which brings almost 10K on its own) really make him more reputable than other users? –  Bruno Jan 12 '12 at 17:27
    
In that rare and very specific case, yes, the metric is broken. However, I still consider it less broken than rep. I've seen too many 1k+ users who have asked a thousand mostly bad questions. –  meagar Jan 12 '12 at 17:48
    
Sure, but if they earned those points mostly with bad questions, this means that readers have upvoted these bad questions. The issue there would unfortunately be with "bad voters", which will make stats meaningless anyway (whether it's for question or answers). This being said, perhaps more distinction between points obtained from questions and those from answers would be better indeed. –  Bruno Jan 12 '12 at 17:59

You must log in to answer this question.

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