You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Branchfeat/t-100a-metrics-met-01-discovery-rate — one task = one branch = one PR (doc 32 §6)
Summary
Implement per-campaign counting metrics in metrics/definitions.py (or metrics/campaign_metrics.py) with exact formulas of doc 19 §3.
Context / motivation
Metric bugs silently invalidate the research claims; formulas are pre-registered in doc 19.
Scope
In scope
Pure functions over RunRecord sequences; MET-03 right-censored at B (returns value + censored flag).
MET-07 with bootstrap CI hook (statistics from T-101a).
Implementation guidance
Exclude confirm-tagged runs from strategy comparison metrics (doc 09 §6). Document every formula in the docstring with the doc reference.
Standards: respect the layering contracts (doc 16 §2) and conventions in doc 34; no hard-coded tunables — thresholds/defaults come from typed settings (doc 27); no magic numbers (ruff PLR2004); public APIs typed (mypy strict) and documented.
Acceptance criteria
Hand-computed examples for each metric (≥ 3 cases each incl. edge cases: no failures, all INVALID).
≥ 90 % coverage.
Tests listed below added and green in CI; coverage of new code meets the gate in doc 18 §3.
Docs, config reference and CHANGELOG Unreleased updated where behaviour, keys or interfaces changed.
Activity
experiment run|report(concurrency-limited) #109