Reputation on SO has been highly discussed. I have been curious how the reputation is distributed. A first glimpse on the userlist resulted in the hypothesis, that reputation is degrading following a power-law to the (reputation-order) rank of the user. This distribution would support the theory expressed by the question whether "High Reputation Attracts Too Many 'Up-Votes'?"

I did some data gathering (privacy aware, not storing any user identifiers, mind you) and got the following graph:

power-law degradation of reputation in SO userlist

On the x-axis you find the rank of the user, starting with 0 (Dijkstra-conform). On the y-axis the reputation value is shown in logarithmic scale. By "rank" I denote the ordinal numbers of the userlist ordered by reputation (order as done by SO)

I am not too good with evaluating data, but that degradation follows something stronger than 1/(a^x) seems quite obvious to me.

What does this say about SO-users? Are we stacking positive votes on top of those who already have a high reputation? Are we 'follow the leader'-types? Do we need more data?

  • Has the graph gone to silicon heaven? – Andrew Grimm Mar 24 '11 at 22:22
  • With power laws, isn't it normal to log the X axis as well? – Andrew Grimm Jul 1 '11 at 14:50

Call me crazy... but wouldn't this data also support the theory that high rep users tend to write good answers, attracting up-votes? Meaning: answers that are readable; accurate; qualified (i.e. with evidence).

I accept that I fall into the "high rep" bracket, so I hope that doesn't sound egotistical. But when I vote, I'm voting for the quality of that question/answer, not the user. If a new user posts a good answer, I'll vote for it.

  • Further evidence for this: I think I've been breaking the rep cap nearly every day since I started... clearly before I was a high rep user. (I'll put a little time into writing a tool to plot that soon...) – Jon Skeet Jul 27 '09 at 10:12
  • Jon, if the growth of your reputation would increase daily, that would be some kind of proof to the theory, but simple growth of reputation would not. Most reputations probably are monotonous and I think that is not only because we tend to vote-up more easily than down, but also because the reputation-system favours positive reputation over negative. – Don Johe Jul 27 '09 at 10:20
  • @Don: My point is that my reputation growth doesn't increase daily... my average is probably slightly higher now than before because I'm still getting votes for old answers, which means I hit the cap earlier (and so more accepted answers count beyond the rep cap). If reputation were all about following the crowd, I would have found it very hard to exceed the rep cap before I became a high rep user. – Jon Skeet Jul 27 '09 at 10:25
  • "f the growth of your reputation would increase daily" - the rep cap largely prevents this – Marc Gravell ModStaff Jul 27 '09 at 10:26
  • This is true, but I've also found a lot of generic answers get upvoted on Stackoverflow just because the answer gives a kind of soundbite answer. Here's an example: stackoverflow.com/questions/1124609/…. There's a lot of what a cynic might say call soundbite answers - answers that read like a paragraph from chicken soup for the soul for programmers. Effective communication seems to help a great deal, which comes from book or article writing which I'm useless at – Chris S Jul 27 '09 at 10:55
  • @Chirs S: 1ST) what you are calling a "soundbite answer" is, by my read, a mildly sarcastic answer to a lazy question from a user who needs to be set straight. There are two answers to date, both along the lines of "write pretty now, optimist later if there is a problem". 2ND) the Q here is about the relation between current reputation and ability of an answer to attract more points--and your example does hold in this context--the 2nd most popular answer (7) is from a user with the lowest reputation compared to the 'winning' answer (1). The post dates are more important in your example. – Stu Thompson Jul 27 '09 at 17:14
  • My highest-voted answers were short and simple. But when I answer something complex that requires research, I may or may not get upvoted at all. It seems strange... – Steven Sudit Aug 3 '10 at 20:14

I've done a couple of related blog posts with my own graphs based on the data dumps which you might want to have a look at:

Two interesting and relevant graphs from those blog posts:

Percent Up Votes (of Total Votes) for Five Reputation Bands + the Top 10 Stack Overflow: Percent Up Votes (of Total Votes) for Five Reputation Bands + the Top 10

Down Votes (x-axis) vs. Up Votes (y-axis) vs. Reputation (z-axis) Stack Overflow: Reputation vs Up Votes vs Down Votes

"Privacy?" Yea, right. We are posting to a Collective Commons lic'd site with public data dumps. The SO team has sanitized to the point where if a users wants to remain anonymous they can. If a users doesn't care, then that is that. Remember, real names and other identifying information is not required/displayed. Accounts are not even a requirement.

In the blog posts are more interesting statistics and graphs, including stuff on specific users.

  • I really do like the last chart. Well, and I finally stumbled upon the database dump. I better use this, next time. – Don Johe Jul 27 '09 at 10:29

You must log in to answer this question.

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