# L6 — Pair / group feasibility notes v0.3 (METADATA ONLY)

**NO correlations · NO coupling · MODEL_CHANGE=NO · density_meta ≠ authorized score**

## Method
| Component | Rule |
|-----------|------|
| Overlap | Inclusive calendar-year intersection **or** congress_period count — unit field required |
| Independence | Source-family heuristic = provenance diversity, **not** statistical independence |
| Frequency | Matched / annualize / congress↔election — NN matching is PROPOSED ONLY |
| density_meta_product | `overlap_years × indep_w × freq_w` only when `overlap_unit=calendar_years_inclusive` |
| Forbidden | Correlations, regression, coupling, outcome/phase labels, score authorization |

## OPP corrections (before → after)
| OPP | Before | After |
|-----|--------|-------|
| OPP7 | 11y | **9y** inclusive; SCA raw 108 months; do not borrow SCE window |
| OPP1 | group 48y with dens 75 | **pair** 75y/dens75; **group** 49y/dens49 (separated) |
| OPP4 | 21 overlap_years | **21 congress_periods** ≠ years; calendar span ~42y separate |
| trust | 68 as coverage | 53 years_with_polls ≠ 68 calendar |

## Outputs
- `pair_feasibility_v0.3.csv`
- `ranked_opportunities_v0.3.json` — curated priorities **separate** from reproducible feasibility pair sort


## Ranking label (v0.3.1)
**SCOUTING_ONLY** — calendar-span × heuristic weights. Not an accepted feasibility ranking.
`observation_coverage_status` remains **UNKNOWN** for levels/diff/lag/rolling/holdout until per-lag complete counts exist.
Exact ICS∩PAGO∩PEXP month intersections are exported under `exact_intersections_v0.3.1.json`.