# v2.1 method analysis

- Model: `2.1.0`
- Profile: `fixed-five-event-response-v2.1`
- Score profile SHA-256: `fba67e0308b51c8c28526456a327e6e31638852f7a73b492e856dda3a7546b11`
- Valid cached state vectors: 325035

## Main result

The v2.1 headline value is, by construction, exactly the **worst event-conditioned initial-RoCoF response margin** among the five fixed score-driving cases. Nordic inertia remains physically present in generator-event RoCoF through post-event kinetic energy, but its normalized 90/150 GWs context curve is no longer a separate per-scenario limiter.

This analysis is descriptive and sensitivity-oriented. It is not prospective validation against an independent security outcome.

## Baseline comparisons

- Worst event-RoCoF identity: `True`
- Pearson vs global inertia context: `0.3715904669852584`
- Spearman vs global inertia context: `0.7234488223377943`
- Pearson vs reference incident: `0.44514023139915715`
- Spearman vs reference incident: `0.7234488223377943`

## Driver frequencies

- 2017: NORDIC_REFERENCE_INCIDENT_CASE=22730, WIND_COMMON_MODE_CASE=672
- 2018: NORDIC_REFERENCE_INCIDENT_CASE=33437, WIND_COMMON_MODE_CASE=1615
- 2019: NORDIC_REFERENCE_INCIDENT_CASE=31313, WIND_COMMON_MODE_CASE=3625
- 2020: NORDIC_REFERENCE_INCIDENT_CASE=25572, WIND_COMMON_MODE_CASE=9374
- 2021: NORDIC_REFERENCE_INCIDENT_CASE=25573, WIND_COMMON_MODE_CASE=9462
- 2022: NORDIC_REFERENCE_INCIDENT_CASE=22494, WIND_COMMON_MODE_CASE=12529
- 2023: NORDIC_REFERENCE_INCIDENT_CASE=21931, WIND_COMMON_MODE_CASE=13087
- 2024: NORDIC_REFERENCE_INCIDENT_CASE=17687, WIND_COMMON_MODE_CASE=16383
- 2025: NORDIC_REFERENCE_INCIDENT_CASE=18144, WIND_COMMON_MODE_CASE=16749
- 2026: NORDIC_REFERENCE_INCIDENT_CASE=12027, WIND_COMMON_MODE_CASE=11045

## Time-series and sensitivity evidence

The machine-readable artifact contains per-year baseline comparisons, tie-aware sampled Kendall tau-b, 1/2.5/5/10% lower-tail Jaccard and directional containment, lower-5% episode overlap under 30/60/120/240 minute recovery gaps, leave-one-case-out ablation, RoCoF-reference/score-anchor/transform sensitivity, and ±20% generic generator-H sensitivity.

## Live JAO worked examples

The JSON artifact contains one complete min-raw-slack CNE example and one PTC example with delivery MTU, backup state, eligible-set hash, RAM/F0/Flow_FB/Fmax, raw/normalized slack, shadow price and reconciliation fields from the current published JAO state.

## Reproduction

Full numerical evidence is in `v2-method-analysis.json`; no prospective event outcome is used to optimize or validate this stable implementation release.