Recently, I'm interested in CF rating and contribution distribution, and I searched this.
I already searched about mysteries of rating distribution (Blog), so I wrote here about contribution distribution. (Thank you for comments)
Method of Searching
- Here is the leaderboard of top contributor.
- I searched the rank of contribution -51, -46, -41, -36,..., -1, 4, 9, 14,..., 164, 169, 174, 179. I searched the number of contributors too.
- Then, I make the distribution graph.
- Each range of the graph is [-inf, -50), [-50, -45), [-45, -40),... , [175, 180), [180, inf]. Each value of x-axis of the graph is the minimum value in range. For example, if the value of axis is 15, this means the range is [15, 20).
- The minumum value of range [-inf, -50) is inf, but in distribution graph this value is -1000.
However, there was some surprising thing about it.
Look at following graph:
Note: The distribution graph has Logarithmic scale. Every time the scale increases by 1, the value doubles (2 times).
The surprising things are following:
- Obviously, the number of people who have contribution [100, 105), [105, 110), [110, 115) is especially larger than [90, 95), [95, 100).
- The number of people decreases greatly from contribution [10, 15) to [15, 20). The difference is approximately 2.5 times.
- The number of people who have contribution [-50, -45) , [-45, -40), and [-40, -35) are nearly the same.
Why these things can be occur?
UPD1: rng_58's contribution became 180, so the graph was extended. Wonderful.
UPD2: The distribution graph is extended because some user's contribution was changed.