I whipped up a graph this morning plotting each user with more than 10 comments.

Neil Butterworth wins my world famous "Quality Volume Commentator of the Year" award. Of the top ten commentators, he has the highest average score.

Number of Comments vs Average Score by User alt text http://lanai.dietpizza.ch/images/stack-overflow-comments-user-total-vs-avg-score-small.png

(Larger image)

Also, I was looking for odd/interesting relationships between comment scores and whatever. The below graph shows 12 comment score bands versus:

  • Average Comment Length
  • Average User Id (vaguely, the average age of an account)
  • Number of Comments

Note that the left side y-axis is logarithmic.

alt text http://geekomatic.ch/image/stack-overflow-comment-score-vs-count-avgUserId-avgLength.png

There most definitely seems to be a relationship between comment length and score--shorter comments score higher. And there is a vague hint that the older SO accounts score higher, but it seems slight.

Any other interpretations? Questions? Comments?

(I've also made an effort to keep these color blind-friendly. Please let me know how I did.)

  • 8
    TheTXI looks like a Brachiosaurus. – random Sep 8 '09 at 11:05
  • 3
    +1 - I love these analysis posts. I wish Jeff would write posts like this on the blog. – Matt Hamilton Sep 8 '09 at 11:26
  • 5
    Need a meta version of this. – TheTXI Sep 8 '09 at 12:25
  • 3
    Can you run this again after eliminating "should be wiki" and "should be community wiki" comments from the pool? Those tend to get upvoted a lot, and might be skewing your numbers. – Bill the Lizard Sep 8 '09 at 12:39
  • @Bill: Good catch. While there are not that many comments that meet this criteria (~600), the #1 user who does this, Niel Butterworth, has 77 comments with an ave vote score 3.3! Everybody else is not even close in either volume or avg score. – Stu Thompson Sep 8 '09 at 13:22
  • 7
    I second the motion that you make a meta version of this graph. Do it or I will braid your dachshunds. – Welbog Sep 8 '09 at 13:27
  • @Stu: Given your earlier analysis involving tag popularity, particularly the 'subjective' tag, I thought there would have been more "should be wiki" comments than that. It sounds like this graph won't change that much (if at all) if you eliminate those. Thanks for checking that out. – Bill the Lizard Sep 8 '09 at 13:32
  • ...also, I know they're not circles, but you should definitely get some cred for the freehand pointers you drew in the first graph. – Bill the Lizard Sep 8 '09 at 13:33
  • 3
    @Billy: I like how TXI's name is nowhere near his dot. Maybe the lines mean something? Jon Skeet doesn't have a line, but TXI has a huge one! Perhaps this represents degree of separation from reality. – Welbog Sep 8 '09 at 13:36
  • +1 for my name tag being so far out there. – TheTXI Sep 8 '09 at 13:36
  • Can I just check - does this graph really say I've written by far the most comments, or am I misreading it? – Jon Skeet Sep 8 '09 at 13:38
  • 6
    Brachiosaurus is the ancient ancestor of ponies, before they traded in girth for cute. – TheTXI Sep 8 '09 at 13:47
  • 1
    No meta data booooo – Welbog Sep 8 '09 at 13:47
  • 1
    Wow. I really wouldn't have expected to have that many comments. I guess a fair number of them are in replies to other comments to my posts, so with the most posts it's a natural correlation... – Jon Skeet Sep 8 '09 at 13:50
  • 2
    @Jon Skeet: If you eliminate all the comments pointing out the errors in the first edition of Head First C# it would probably bring you down to the rest of the pack. – Bill the Lizard Sep 8 '09 at 14:36

I'm curious how the first graph changes if you do exclude comments with no votes, so you only count comments with at least one vote.

(Edit by Stu)

Number of User Comments with Score >= 1 vs Average Score of Votes with score >= 1 for users with at least 10 comments alt text http://lanai.dietpizza.ch/images/stack-overflow-comments-user-total-vs-avg-score-with-at-lease-one-up-vote-small.png (Larger image)

  • I've molested your post, Joel. Hope you don't mind. – Stu Thompson Sep 8 '09 at 17:39
  • Awww! I'm not in that one. :( – gnostradamus Sep 8 '09 at 17:48
  • That's cause I am a lazy git. I'll try to get it on tonight, and anyone else who whines in the next few hours. – Stu Thompson Sep 8 '09 at 18:31
  • 2
    It looks like only about 10% of my comments get upvoted. I guess I need to start swearing like a drunken sailor. – Bill the Lizard Sep 8 '09 at 19:03
  • 2
    Kyle, gnovice, any other whiners that may follow: I am tired and grumpy and going to bed. You'll get your narcissistic labels in the morning. – Stu Thompson Sep 8 '09 at 19:43
  • OK, kids--you've been added to the graph. – Stu Thompson Sep 9 '09 at 11:54
  • 1
    Hurray! Bill and I are chart buddies! – gnostradamus Sep 9 '09 at 13:24

Seems to be trending downwards except for two spikes. Those interest me at least the most. Since the graph is log, the spike in long comments is quite interesting. And I'm wondering if its welbog?

  • Average comment length is on the right-side, green y-axis. That is not logarithmic. Note that once the 'score band' gets above 50 or so, the number of comments generally shrinks to the point of being statistically very vague. E.g.: that spike has only ~12 data points. – Stu Thompson Sep 8 '09 at 11:32
  • Moreover I don't comment as much on SO. Usually over there I do more editing and answering and whining than commenting. – Welbog Sep 8 '09 at 13:13
  • 2
    Now if this graph were to be made for Meta... – Welbog Sep 8 '09 at 13:14
  • @Welbog: You would certainly win in the curse per comment (C/C) factor. – TheTXI Sep 8 '09 at 13:25
  • I've added Welblog to the graph for completeness' sake. – Stu Thompson Sep 8 '09 at 14:10
  • I'm standing outside, waiting in line to get into the blob. I need to get my name on the list! – Bill the Lizard Sep 8 '09 at 14:34
  • 3
    @Billy: I can get you in if you promise to buy me a donut whenever I ask you for one, no questions asked. Please note that this might result in me calling you up at 3 in the morning because I have a massive craving for a powdered lemon jelly donut. Or maybe I'll call you up in the middle of a meeting because I need my esophagus to experience the righteous flow of a masticated glazed donut. And I would use those terms on your conference room's speaker phone, too. Your boss will think you're into some weird donut-and-ingestion-based fetishes. As opposed to the lizard-based ones you're into now. – Welbog Sep 8 '09 at 14:43

I am utterly incapable of parsing the last graph, because the score is on the horizontal axis (input). Not to mention shoving three graphs into one, when they don't even have the same scales!

Wouldn't it make more sense to graph the average score of comments with length N instead of the average length of comments with score N? I can control the length of my comments, but I can't control my comments' score (at least not in the same direct sense).

  • Understand that these puppies take time to produce. Three graphs is significantly more effort. Different queries to produce different axises is much more effort. (More effort == fewer graphs.) You are more than welcome to do them yourself in any fashion you like--the data dump is public. – Stu Thompson Sep 8 '09 at 19:25

Are short comments likely to be pithy jokes or the oft used "why do you want to do this instead of X"

Is your first graph X: number of comments, Y: Average score?

Have you considered doing the graph for comments only created since the ranking was available (many old comments simply don't get any love in this regard)

  • I've added a title. Thanks. Yes, I had. But I couldn't remember when those came about and just wanted to finish the darn thing. It should not be too much work to do again, if you know the date. – Stu Thompson Sep 8 '09 at 12:33

You must log in to answer this question.

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