$ research --validate --model regime-v6
LOADING…
$ research --audit --explainquestion: does our regime-switching model actually work on out-of-sample data?
evidence: R² per slot · p-values · correlation stability · FRED provenance.
audience: macro researchers, enterprise quants, skeptical readers.
last_run: 2026-05-22 up_to_date
$ audit --section regime --verbose
1. REGIME MODEL AUDIT
Current regime:RECOVERY
Regime age:11M
Total transitions observed:8
REGIME DURATION SUMMARY
| REGIME | COUNT | AVG DURATION |
|---|---|---|
| PEAK | 2 | 11.5M |
| CONTRACTION | 2 | 3.0M |
| TROUGH | 1 | 11.0M |
| RECOVERY | 2 | 7.0M |
| EXPANSION | 1 | 7.0M |
TRANSITION HISTORY
| DATE | FROM | TO | DURATION |
|---|---|---|---|
| 2021-11 | EXPANSION | PEAK | 18M |
| 2022-01 | PEAK | CONTRACTION | 2M |
| 2022-12 | CONTRACTION | TROUGH | 11M |
| 2023-03 | TROUGH | RECOVERY | 3M |
| 2023-10 | RECOVERY | EXPANSION | 7M |
| 2024-03 | EXPANSION | PEAK | 5M |
| 2024-07 | PEAK | CONTRACTION | 4M |
| 2025-06 | CONTRACTION | RECOVERY | 11M |
$ audit --section cycle-fit
2. CYCLE FIT QUALITY
| ASSET | R-SQUARED | PERIOD | AMPLITUDE | LEAD/LAG | QUALITY |
|---|---|---|---|---|---|
| BTC | 0.420 | 48M | 1.82 | 0M | MODERATE |
| ETH | 0.380 | 48M | 2.14 | +1M | MODERATE |
| SPX | 0.310 | 42M | 0.67 | -3M | MODERATE |
| Gold | 0.180 | 60M | 0.44 | -6M | LOW |
| DXY | 0.240 | 38M | 0.52 | -2M | MODERATE |
| VIX | 0.080LOW | 24M | 1.35 | 0M | LOW |
| Net S-D | 0.610 | 46M | 1.00 | 0M | GOOD |
R-squared measures the proportion of variance explained by the cycle model. Lead/lag is relative to Net S-D composite (negative = leads). Assets with R-squared below 0.10 are flagged as LOW CONFIDENCE.
$ audit --section correlations --windows 30,90,365
3. CORRELATION STABILITY
| PAIR | 30D | 90D | 365D | STATUS |
|---|---|---|---|---|
| BTC/SPX | 0.72 | 0.65 | 0.58 | STABLE |
| BTC/ETH | 0.94 | 0.91 | 0.88 | STABLE |
| BTC/Gold | 0.31 | 0.22 | 0.15 | STABLE |
| BTC/DXY | -0.48 | -0.42 | -0.35 | STABLE |
| BTC/VIX | -0.56 | -0.51 | -0.38 | STABLE |
| SPX/VIX | -0.82 | -0.79 | -0.75 | STABLE |
| Gold/DXY | -0.61 | -0.55 | -0.48 | STABLE |
| ETH/SPX | 0.68 | 0.61 | 0.52 | STABLE |
$ audit --section significance --alpha 0.05
4. STATISTICAL SIGNIFICANCE
| SIGNAL | P-VALUE | 95% CI | N | STATUS |
|---|---|---|---|---|
| BTC | 0.0030 | [0.62, 0.84] | 120 | SIGNIFICANT |
| ETH | 0.0080 | [0.55, 0.79] | 96 | SIGNIFICANT |
| SPX | 0.0010 | [0.71, 0.89] | 240 | SIGNIFICANT |
| GOLD | 0.0420 | [0.51, 0.68] | 240 | SIGNIFICANT |
| DXY | 0.0150 | [0.58, 0.76] | 240 | SIGNIFICANT |
| VIX | 0.0870 | [0.44, 0.62] | 240 | MARGINAL |
P-values test the null hypothesis that the signal has no predictive relationship with the target variable. Green (p < 0.05) = statistically significant at 95% confidence. Amber (p < 0.10) = marginal. Red (p >= 0.10) = not significant.
$ audit --section provenance --list-series
5. DATA PROVENANCE
Total FRED series:12
Signal count:6
Data source:Static JSON (fallback)
FRED SERIES INVENTORY
| SERIES ID | NAME | FREQUENCY |
|---|---|---|
| WALCL | Fed Balance Sheet | Weekly |
| RRPONTSYD | Reverse Repo (ON RRP) | Daily |
| WM2NS | M2 Money Supply | Monthly |
| BOGMBASE | Monetary Base | Biweekly |
| DGS10 | 10-Year Treasury Yield | Daily |
| DGS2 | 2-Year Treasury Yield | Daily |
| VIXCLS | VIX | Daily |
| SP500 | S&P 500 | Daily |
| GOLDAMGBD228NLBM | Gold Price (London) | Daily |
| DTWEXBGS | Trade Weighted Dollar (DXY) | Daily |
| CPIAUCSL | CPI (All Urban) | Monthly |
| UNRATE | Unemployment Rate | Monthly |
SIGNAL LAST UPDATED
| KEY | SOURCE | LAST UPDATED | N OBS |
|---|---|---|---|
| BTC | FRED: CBBTCUSD | 2026-03 | 120 |
| ETH | CoinGecko | 2026-03 | 96 |
| SPX | FRED: SP500 | 2026-03 | 240 |
| GOLD | FRED: GOLDAMGBD228NLBM | 2026-03 | 240 |
| DXY | FRED: DTWEXBGS | 2026-03 | 240 |
| VIX | FRED: VIXCLS | 2026-03 | 240 |