MERGED.

Method

What counts, and
what doesn’t.

A ranking nobody can interrogate is a ranking nobody should trust. Here is the whole formula, including the parts of it that are wrong.

The five parts

1,000 points available, distributed on purpose

Why nothing scales linearly

Each part runs through a saturating curve rather than a straight line. The distance between zero and fifty upstream patches is a career change; the distance between five hundred and five hundred fifty is noise. A linear score would say those gaps are equal, so the curve says otherwise.

score = Σ weight × (1 − e^(−value / k))

The ladder

Tier names are borrowed from the roles real projects give their own people, rather than invented metals.

Where this is wrong

Every ranking encodes a bias. These are the ones known about:

Data comes from GitHub’s public API and public contribution graphs, read without an account.

+ Add your work