IS/OOS 2016-01-01, sqrt(365). Each candidate is audited for independence: residual Sharpe vs EW equity market + correlation to validated momentum (EqMom) and to the market. A real NEW edge = positive Sharpe AND low corr to both.
A. Treasury / bond time-series momentum (long if 12-mo return>0)
| Asset | OOS Sharpe | p | Net@5 | Resid vs Mkt | Beta | Corr EqMom | Corr Mkt |
|---|
| TLT | 0.14 | 0.362 | 0.13 | -0.00 | -0.19 | -0.26 | -0.35 |
| IEF | 0.32 | 0.195 | 0.29 | -0.00 | -0.09 | -0.27 | -0.34 |
| SHY | 1.90 | 0.000 | 1.84 | -0.00 | -0.02 | -0.23 | -0.27 |
| IEI | 0.84 | 0.011 | 0.80 | 0.00 | -0.05 | -0.25 | -0.31 |
| AGG | 0.62 | 0.048 | 0.59 | -0.00 | -0.01 | -0.05 | -0.07 |
| LQD | 0.51 | 0.080 | 0.49 | 0.00 | 0.02 | 0.07 | 0.08 |
| HYG | 0.79 | 0.016 | 0.78 | 0.00 | 0.16 | 0.49 | 0.47 |
| TIP | 0.42 | 0.123 | 0.39 | -0.00 | -0.03 | -0.10 | -0.13 |
B. Commodity cross-sectional momentum (top-2 of 6)
| Factor | OOS Sharpe | p | Net@10 | Resid vs Mkt | Corr EqMom | Corr Mkt |
|---|
| CommodXS(2) | 0.78 | 0.016 | 0.76 | 0.00 | 0.23 | 0.17 |
C. FX cross-sectional momentum (top-1 of 5 vs USD)
| Factor | OOS Sharpe | p | Net@5 | Resid vs Mkt | Corr EqMom | Corr Mkt |
|---|
| FXXS(1) | 0.21 | 0.284 | 0.08 | -0.00 | 0.13 | 0.05 |
D. Curve-steepener timing (long TLT when 10y-3m slope rising)
| Factor | OOS Sharpe | p | Resid vs Mkt | Corr EqMom | Corr Mkt |
|---|
| Steepener | -0.08 | 0.584 | 0.00 | -0.21 | -0.24 |
E. Gold flight-to-quality (long TLT when VIX rising)
| Factor | OOS Sharpe | p | Resid vs Mkt | Corr EqMom | Corr Mkt |
|---|
| VIX-TLT | 0.06 | 0.433 | 0.00 | -0.27 | -0.31 |
Reading — IMPORTANT METHODOLOGY NOTE
For NEW asset classes the right independence test is correlation to equities,
NOT residual-Sharpe-vs-market. Regressing a low-vol bond (vol ~1-4%) on the
high-vol equity market (~28%) yields a degenerate ~0 residual — that does NOT
mean "no edge", it means the asset lives on a different vol curve. The
correlation column is the real signal: Treasuries show negative correlation
to equities (true diversification); commodities/FX show low positive.
Characterization (full history, incl. 2008/2022)
| Asset | Sharpe | CAGR | annVol | maxDD | corr Mkt |
|---|
| SHY (1-3y) | 1.75 | 2.7% | 1.5% | -2.2% | -0.27 |
| IEI (3-7y) | 0.88 | 3.4% | 3.9% | -6.0% | -0.31 |
| AGG (agg) | 0.84 | 3.7% | 4.5% | -9.6% | -0.07 |
| TIP (TIPS) | 0.76 | 4.2% | 5.6% | -13.0% | -0.13 |
| CommodXS(2) | 0.31 | 4.7% | 31.1% | -73.1% | 0.17 |
(Treasury Sharpe is a low-vol artifact — they are bond CARRY, not alpha, but
genuinely independent, low-DD return streams. Commodities are independent but
extremely crashy -> satellite-only.)
Diversification benefit (no leverage)
| Book | Sharpe | CAGR | annVol | maxDD |
|---|
| EqMom only | 1.10 | 31.0% | 28.1% | -54.9% |
| EqMom 70% + SHY 30% | 1.15 | 23.2% | 19.9% | -40.6% |
| EqMom 50% + SHY 25% + Commod 25% | 1.06 | 18.8% | 17.7% | -39.5% |
Adding negative-correlation Treasuries RAISES risk-adjusted return and nearly
halves tail risk. Commodities add independence but a negative Sharpe
contribution (too crashy), so a Treasury sleeve is the better diversifier.
Verdict — ROUND 4 FOUND genuine new, independent sleeves
- Short/intermediate Treasuries (SHY, IEI, AGG, TIP) = real, independent
diversifiers: NEGATIVE equity correlation (-0.07 to -0.31), significant
positive Sharpe, tiny drawdowns. Bond carry, not alpha, but the first
genuinely non-beta return stream we have found. Best used to cut book tail
risk (EqMom+SHY: Sharpe up, DD -55% -> -41%).
- Commodity cross-sectional momentum = independent (corr ~0.2) but high-vol
and -73% DD -> small satellite only; net negative Sharpe contribution.
- REJECTED: FX cross-sectional (p=0.28), curve-steepener timing (Sharpe -0.08),
VIX-TLT timing (Sharpe 0.06). HYG is equity-like (corr 0.47) -> not a diversifier.
This is the payout of getting NEW data: a multi-asset book (equity momentum +
crypto gate60 + Treasury diversifier) is now genuinely diversified, not just
two flavors of beta.