scoring algorithm · pinned per scorecard

per-check-mean-v1

Per-check-identity mean, introduced in v0.3.0-draft to fix flat-pool's page-count dependence.

Introduced in the v0.3.0-draft to fix the page-count dependence in flat-pool-v1:

score = round(mean({ passed/applicable for each check_id where applicable > 0 }))

Each distinct check identity contributes one observation, regardless of how many pages it fires on. Site-wide checks and per-page checks sit on equal footing per check id; the cap a site is crawled at no longer changes the weighting of its site-wide signals. For 1-page audits the score is identical to flat-pool-v1; for site-mode audits inflated cap-hit scores come down toward a more representative number (e.g. posthog-docs 97 → 41 on a 500-page audit, where the 97 was masking a long tail of failing per-page checks via high na rates).

See all scoring methodologies or browse scorecards to see which methodology each version pins.