3

The Stack Overflow Leagues page displays the count of users per reputation segment. Recent snapshot (with percentages added for perspective):

Total Rep      Users  Portion
---------  ---------  -------
100,000+         397    0.01%  
50,000+         1204    0.02%
25,000+         3287    0.05%
10,000+        10085    0.17%
5,000+         22176    0.37%
3,000+         37928    0.63%
2,000+         56428    0.94%
1,000+        104678    1.75%
500+          184194    3.08%
200+          311747    5.21%
1+           5987285  

The breakdown stops at 200 rep, below which the vast majority (95%) of our users lie. There's even a footnote:

users with less than 200 reputation are not tracked in the leagues

Show more granularity for the masses!

Suggestion: Add segments for 100, 50 and 15.

The thresholds of 50 and 15 are noteworthy, because they bestow the privileges of commenting and up-voting respectively, which in my experience are the most talked about, sought after and problematic privileges. Adding 100 is an obvious logarithmic step between 200 and 50.

Here's the result of a query that includes those segments:

Total Rep      Users  Portion
---------  ---------  -------
100,000+         397    0.01%  
50,000+         1204    0.02%
25,000+         3287    0.05%
10,000+        10085    0.17%
5,000+         22176    0.37%
3,000+         37928    0.63%
2,000+         56428    0.94%
1,000+        104678    1.75%
500+          184194    3.08%
200+          311747    5.21%
100+          494425    8.25%  <-- new segment
50+           716959   11.97%  <-- new segment (comment privilege)
15+          1313555   21.94%  <-- new segment (up-vote privilege)
1+           5987285  

Graphing these numbers shows a smooth curve across the entire dataset, as shown by this log version of the above query:

logarithmic graph

so it's not as if the numbers get strange at lower reputation levels.

Is there a compelling reason there is not finer granularity?
Disk is cheap. CPU is cheap. Why not do it?

  • I expect this hasn't been done yet because the bucketing comes as a side-effect of the user-by-user stats, which would be rather more expensive to compute below 200 rep. Still, DRY aside, this shouldn't be especially difficult to implement. – Nathan Tuggy Sep 8 '16 at 23:06

You must log in to answer this question.

Browse other questions tagged .