7000.0.0a16
Alpha · Released Nov 5, 2026 · Changeset 0745a91df592 · 7 release notes
77build scorehigher = better
050100
Upgrade risk+33Net cleanup+25Live debt+20
cohort: alpha (45)
How this number is computed
Composite = sum of sub_score × weight across six metrics. Sub-scores are percentile-rank(log1p(min(raw, p95))) within the alpha cohort (45 releases). Sub-scores for lower-is-better metrics are flipped to 100 − percentile.
| Sub-score | Formula | Raw | Clip | %ile | Weight | Score | Contrib |
|---|---|---|---|---|---|---|---|
| Breaking surface | (breaking_change + api_change) / total_notes | 0.060 | 0.060 | 48 | 25% | 52 | 13.1 |
| Blocker rate | blockers / total_notes | 0 | 0 | 3 | 15% | 97 | 14.5 |
| Mobile risk | mobile_known_issues / total_notes | 7.8e-4 | 7.8e-4 | 50 | 10% | 50 | 5.0 |
| Fix density | fixes / total_notes | 0.758 | 0.758 | 81 | 15% | 81 | 12.2 |
| Net-fix delta | ((fixes − known_issues) − prior.(fixes − known_issues)) / total_notes | 0.288 | 0.288 | 85 | 15% | 85 | 12.8 |
| Known-issue debt | known_issues / total_notes | 2.7e-3 | 2.7e-3 | 1 | 20% | 99 | 19.8 |
| Composite | 77 | ||||||
* = raw value was clipped at the cohort 95th percentile to prevent outliers from compressing everyone else's scores. Sub-scores marked "med" substitute the cohort median where the input was missing (e.g. first patch on a minor line has no prior patch for net-fix delta).