Hacker News Reader

Comment by pg | Hacker News Reader

    > (sum [karma _] (users)) 
    1632636

This may be a slight undercount, because users are lazily loaded. But most with significant karma would be loaded at this point.

View on HN

So the user with the highest karma has ~2.6% of the total... interesting.

That is assuming pg is included in the count. I haven't looked at the code so I'm not sure if pg is classified as a user or something else.

Some time ago I had a hunch that karma followed a powerlaw, so I mapped the leaderboard out in excel - and sure enough it looked exactly like a powerlaw graph.

The curve I fit to the leader board is,

  karma = 25343 * rank^(-.465)

Which has an R-squared value of 0.9791.

Unfortunately, this isn't a good enough fit to meaningfully extrapolate very far.

27075/1632636 = 1.65%. Am I missing something?

yes: http://news.ycombinator.com/user?id=pg

Ah...

Slightly off-topic, but it's awesome that you post code along with the answer when asked these types of questions.

It would be a slight undercount if karma didn't follow a power-law. How many users are there again? How many were loaded when you ran that?

13527 of 43817 accounts are currently loaded.

What do you mean loaded? You mean in order to filter spam, trolls, or bots?

Users and items (stories, comments) are lazily loaded from disk to make restarts faster.

I did a few quick calculations, most of which were uninteresting. The only thing that's remotely notable is:

0.07% of the users (the top 100) account for 34.7% of the karma.

Numbers in easily consumable python form:

http://tire.less.ly/hacking/junk/karma.py

Reply on news.ycombinator.com