# Misery Index v3 model audit

## Verdict

v3 passes the predeclared construct checks. It no longer turns postseason losses into misery points, does not use inferred pre-1999 round labels, and gives recent success a separate 40-point channel. The frozen evidence layer and scoring specification reproduce the result offline.

## What changed from v2

1. Near-miss burden was removed from the headline. A deep run now supplies relief through wins; its eventual loss is not an additional penalty.
2. The long-run record is published as Legacy Burden /60, while the most recent five completed seasons determine Current Suffering /40.
3. Pre–Super Bowl titles still count, at half the value of an equally old Super Bowl-era title.
4. Playoff-round labels are not inputs. The Dallas 7-versus-8 v2 boundary therefore cannot affect a v3 score.
5. The score is annual. It changes only after a completed season and postseason are frozen.

The v2-to-v3 rank-order Spearman correlation is 0.941: related, but materially different.

## Detroit construct result

At the 2025 cutoff Detroit is rank 8 of 32 with 72.61. Its legacy is 47.53/60 (Severe), while current suffering is 25.07/40 (Severe).

From 2020 to 2025 Detroit's headline changed by -7.05: legacy moved +0.25, but current suffering moved -7.30. The current improvement dominates the slow aging of old title relief.

Recorded Detroit headline-tier transitions: 1930 Historic → 1931 Severe; 1934 Severe → 1935 Afflicted; 1939 Afflicted → 1940 Severe; 1951 Severe → 1952 Afflicted; 1952 Afflicted → 1953 Ordinary; 1957 Ordinary → 1958 Afflicted; 1981 Afflicted → 1982 Severe; 1982 Severe → 1983 Afflicted; 1983 Afflicted → 1984 Severe; 2008 Severe → 2009 Historic; 2010 Historic → 2011 Severe; 2020 Severe → 2021 Historic; 2022 Historic → 2023 Severe.

## Distribution (diagnostic, not calibration)

- Headline tiers: Afflicted 7, Historic 1, Ordinary 11, Severe 13.
- Legacy tiers: Afflicted 4, Historic 1, Ordinary 4, Severe 12, Untroubled 11.
- Current states: Crisis 7, Severe 19, Stable 1, Strained 5.

Tier occupancy was observed only after the specification was locked. Empty or crowded tiers are not grounds for retuning.

## Largest v2 rank-number movements

| Franchise | v2 | v3 | Change | v2 score | v3 score |
|---|---:|---:|---:|---:|---:|
| Detroit Lions | 2 | 8 | +6 | 79.12 | 72.61 |
| Kansas City Chiefs | 25 | 31 | +6 | 26.85 | 26.60 |
| Carolina Panthers | 10 | 5 | -5 | 68.21 | 76.12 |
| Buffalo Bills | 12 | 17 | +5 | 63.82 | 54.67 |
| Philadelphia Eagles | 23 | 28 | +5 | 31.66 | 29.44 |
| New York Jets | 8 | 4 | -4 | 70.08 | 76.33 |
| Cincinnati Bengals | 5 | 9 | +4 | 73.39 | 71.26 |
| New Orleans Saints | 18 | 14 | -4 | 55.42 | 60.90 |

## Known limits

- The 60/40 layer split, half-lives, five-season window, and postseason relief ladder are transparent value judgments, not fitted estimates.
- Historical playoff-win rate does not adjust every season for changing playoff-field size. Sensitivity tests the fixed relief benchmark, but a future opportunity-adjusted model would be a distinct specification.
- Franchise continuity follows the frozen registry; alternate fan-lineage definitions can legitimately produce different histories.
- The index is an outcome instrument. Ownership, off-field scandal, attendance, and subjective heartbreak are outside its headline construct.

## Validation artifacts

- `V3_SPEC.md` — frozen formulas and behavioral checks
- `V3_PRECOMPUTE_LOCK.json` — pre-score hashes
- `test_misery_index_v3.py` — executable construct and regression checks
- `V3_SENSITIVITY.md` and `sensitivity_results_v3.csv` — parameter dependence
- `data/WOLFRAM_VALIDATION.md` — independent source arithmetic
- `V3_WOLFRAM_VALIDATION.md` — independent v3 arithmetic check