Attribution · Live · as of 2026-05-15
Attribution (monthly)
5-factor rolling OLS decomposition (60-day window). Real coefficients computed daily from FRED + yfinance data.
Factor exposures (rolling 60d)
| Factor | proxy | Base | Bull | Bear |
|---|---|---|---|---|
| Rates duration β | Δ FRED DGS10 | -0.0460 | -0.0525 | -0.0485 |
| USD β | % FRED DTWEXBGS | -0.9560 | -1.2981 | -0.6093 |
| Credit spread β | Δ FRED BAMLH0A0HYM2 | -0.0034 | -0.0042 | -0.0023 |
| Equity vol β | Δ FRED VIXCLS | -0.0002 | -0.0003 | -0.0001 |
| Commodity β | % DBC ETF (DBC) | +0.1355 | +0.1755 | +0.1050 |
| α (daily) | intercept | 0.73bp | 1.46bp | -1.01bp |
| R² | 0.468 | 0.450 | 0.514 |
Method: rolling 60d OLS of daily group returns on Δ factor levels (or % returns for price-level factors).
Interpretation
| Factor | Proxy | Interpretation |
|---|---|---|
| Rates duration β | 10Y UST yield daily change | positive = duration exposure (TLT / CGB) |
| USD β | DXY daily change | positive = benefits from strong USD (US equity / UST) |
| Credit spread β | HYG–LQD daily delta | positive = benefits from spread tightening |
| Equity vol β | VIX daily change | negative = punished in risk-off (typical) |
| Commodity β | Bloomberg commodity index | positive = commodity + gold exposure |
On α (residual)
Daily α is the regression intercept. With 114 observations and only ~3 closed decisions in the log, no inferential claim about α is made — per whitepaper §5.3, residual is treated as a disclosure artefact until ≥6 closed decisions accumulate.
Method caveats
- α is daily and statistically indistinguishable from 0 with <6mo sample
- Coefficients update daily as the rolling window moves
- Commodity proxy uses DBC ETF (BCOM index not on yfinance)
- v1 does not bootstrap confidence intervals