methodology.
Every score is computed from public data by one published formula — percentiles all the way down, no hand-tuned caps, no opaque model. This page renders directly from the same formula definition the scoring engine executes, so what you read here is what runs.
the four steps
- 1metric percentiles
Every raw metric becomes its percentile rank over the entire pool, claimed and unclaimed alike — the share of the pool with a smaller value, ties counting half.
- 2dimension scores
Each dimension is the plain average of its metric percentiles.
- 3source sub-scores
A source’s sub-score is the weighted sum of its dimension scores, using the published weights below.
- 4headline
Sub-scores are sorted descending, weighted 1, ½, ¼, …, and summed into a composite; the headline 0–100 is the composite’s percentile rank over the whole pool.
the formula, drawn
Line thickness is the weight. This diagram is generated from the formula definition — when the formula changes, it redraws itself.
worked example
Every number below is computed by the same engine that scores real profiles — check the arithmetic yourself.
github — dimensions & metrics
| dimension | weight | metrics |
|---|---|---|
| output code shipped — and accepted by others | 40% | commits · last 12 months pull requests merged into others’ repos · last 12 months |
| adoption others building on your work | 25% | stars across owned repos |
| consistency showing up week after week | 20% | active weeks · of the last 52 |
| influence reach and trust in the community | 15% | followers code reviews on others’ pull requests · last 12 months |
versioning
Formula versions are frozen: recalibrating anything — weights, metrics, scales — is a new version, and every score snapshot records the version that produced it. History is never silently rewritten.
Seeded profiles carry provisional scores until their first real computation; claiming a profile computes it from live data immediately.
/slashscore · public data only · no PII stored