9
$\begingroup$

I created a small query to produce a list of number of votes per reputation among +1k users (49 in total). This is, in my view, one of the many indicator of relative contribution to the site. Voting is important primarily because it works as a signaling process conveying information about quality of Q&A. They are the equivalent of prices in real economies, of which importance we as economists are very aware. It also acts as a reward for contributing users asking good questions and providing good answers.

As I see it, a user with a low vote/reputation ratio is one which is more towards the leeching side of things, whereas one with a high ratio is more towards the seeding side of things, if I might borrow the torrent terminology. You might of course differ.

The full ranking is in the page above (after you run the query). Below I copy top and bottom. VotesPerRep is in percentage terms.

  • Top 15:

enter image description here

  • Bottom 15:

enter image description here

I am of course not going to point out names, except for congratulating those with high raking like denesp and Kitsune (I have not interacted with cc7868 and Jamzy). There is however a massive disparity among the group, which if anything is negatively correlated with reputation. This might be due to some outliers though:

enter image description here

A more refined analysis would remove those users that are highly inactive (maybe those which have not provided an answer or question e.g in the last six months), but that is left to whoever wants to play around with the query.

$\endgroup$
15
  • 6
    $\begingroup$ I wondered about how frequently people vote, because there are some questions with two answers where I am the only one upvoting the question. Seems strange that so many people think an answer is needed yet not one thinks the question is good enough to upvote. Thank you for the data. $\endgroup$
    – Giskard
    Oct 13, 2017 at 20:23
  • $\begingroup$ @KitsuneCavalry You have the 6th highest number of downvotes?? I used to think you were nice! $\endgroup$
    – Giskard
    Oct 13, 2017 at 20:27
  • 3
    $\begingroup$ More voting is good, but people should also take the time to read and think about posts before voting. I'd rather have few votes that are highly informative than a voting system where the noise is bigger than the signal. Looking at voting patterns on this site, I sometimes feel like we are closer to the latter end of the spectrum. $\endgroup$
    – Ubiquitous
    Oct 13, 2017 at 21:41
  • 1
    $\begingroup$ @denesp By night, I am the ninja assassin, slicing away reputation 2 points at a time. You have brought my secret into the light, and for that, I must exact my vengeance by winning the Nobel prize in 60 years and calling you a baloney-head in my speech. Your fate has been decided xddd $\endgroup$
    – Kitsune Cavalry Mod
    Oct 14, 2017 at 2:36
  • $\begingroup$ Serious comments: I'm pleasantly surprised my voting is relatively high. I'll gladly take that to the bank. As for the huge clump of points down there on the graph, maybe you wanna make one of the axes a log scale. $\endgroup$
    – Kitsune Cavalry Mod
    Oct 14, 2017 at 2:38
  • 1
    $\begingroup$ @Ubiquitous I entirely agree. I hope I am not being missunderstood as calling people to vote just for the sake of it (and well, the corresponding badges). But I do believe voting is the other side of the coin to answering and asking. Otherwise, we end up being simply another Quora or Yahoo Answers. $\endgroup$
    – luchonacho
    Oct 14, 2017 at 9:07
  • $\begingroup$ @KitsuneCavalry Good suggestion. I am normally confused with log-scale graphs of cross-section data, but will play around with the data once I have more time. $\endgroup$
    – luchonacho
    Oct 14, 2017 at 9:11
  • 1
    $\begingroup$ @luchonacho No accusation intended. My comment is just a polite reminder to people to think before they vote. $\endgroup$
    – Ubiquitous
    Oct 14, 2017 at 9:16
  • 2
    $\begingroup$ Nice to see I made it in the top 15 $\endgroup$
    – EconJohn Mod
    Oct 15, 2017 at 1:58
  • $\begingroup$ Is this the number of votes for or by each user? $\endgroup$
    – user18
    Oct 16, 2017 at 1:44
  • $\begingroup$ @KennyLJ By each user. $\endgroup$
    – luchonacho
    Oct 16, 2017 at 8:05
  • $\begingroup$ @luchonacho I'm not sure I follow the reasoning here. Let's look at cc7768, for example. The upvotes were made by cc7768, or by other users? Assuming the same for the downvotes? $\endgroup$
    – An old man in the sea.
    Oct 17, 2017 at 14:27
  • $\begingroup$ @Anoldmaninthesea. Votes are made by the person. So in your case, you have made 90 votes up and 5 votes down, as it is also shown here (photo above not up-to-date of course). $\endgroup$
    – luchonacho
    Oct 17, 2017 at 14:40
  • $\begingroup$ Lucho, so are we to vote in questions, even if we don't have any interest in them, or know nothing about(my case)? I see a case of a user, who probably is someone who has most of the best answers in this site, and still ended up in the bottom. This metric hardly measures more than a small part of 'forum engagement' by the users. It may neglect people who have contributed immensely by answering numerous questions, but just didn't upvote the question, or other answers. Why not divide upvotes*(answers+questions)/reputation? $\endgroup$
    – An old man in the sea.
    Oct 17, 2017 at 16:37
  • $\begingroup$ @luchonacho ok ;) $\endgroup$
    – An old man in the sea.
    Oct 17, 2017 at 19:29

2 Answers 2

Reset to default
4
$\begingroup$

I find a methodological problem here: I would not use "reputation" but rather "number of questions posted + answers replied": Reputation may correlate with the latter, but it also has an element of "popularity".

But it appears you want to compare different aspects of activity of each user. So it would be better to calculate "for every post (except comments) how many votes".

The ranking won't necessarily be much different, but it would be more sound methodologically, free of "popularity bias".

(Also, you won't need to divide reputation by 100 as you now do).

$\endgroup$
2
  • $\begingroup$ Yes, that is an interesting comparison too, and probably complementary to the one I used. Another one is to look at number of months of activity or so. I am not trying to imply that such "is the best one". But imo it has the "nice" property that you are comparing apples with apples (as votes give/take reputation). Notice the division by 100 is just to transform the VotesPerRep into a percentage. $\endgroup$
    – luchonacho
    Oct 14, 2017 at 16:27
  • $\begingroup$ @luchonacho Reputation is predominantly determined by votes that count for 10 rep points. So, in your votes framework, I think you should divide reputation by 10 and leave votes as is. Then you would obtain the ratio "for each vote received how many votes given". That's indeed another meaningful metric. $\endgroup$
    – Alecos Papadopoulos
    Oct 14, 2017 at 17:25
1
$\begingroup$

At the instance of Lucho, I'm transposing my comment into an answer, even though I'm not sure if it'll bring anything new to the table.

«Lucho, so are we to vote in questions, even if we don't have any interest in them, or know nothing about(my case)? I see a case of a user, who probably is someone who has most of the best answers in this site, and still ended up in the bottom. This metric hardly measures more than a small part of 'forum engagement' by the users. It may neglect people who have contributed immensely by answering numerous questions, but just didn't upvote the question, or other answers. Why not upvotes*(answers+questions)/reputation?»

$\endgroup$
1
  • $\begingroup$ I agree that voting should be informed. But the number of questions and answers are an indirect indication of interest and/or knowledge of certain areas, so for example in your case, asking a high number of questions puts you in the capacity to provide a lot of votes to the answers you get (either good or bad). I agree that users with very popular answers and questions will have a bias in the measure I am using. I do not claim that my measure is the only one available, nor the most complete. A more comprehensive analysis of participations is clearly multidimensional. $\endgroup$
    – luchonacho
    Oct 18, 2017 at 14:47

You must log in to answer this question.

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