Weekly W-FRI SPY, signals lagged 1 week (PIT). Overlay = long/flat with 10bps/switch. Placebo = random gate, matched exposure, 2000 draws; corr p = block-permutation (block 4w). Required per reports/placebo_audit.md.
A. Fed net liquidity (WALCL - TGA - RRP)
| n | corr(4w change-z, next-wk ret) | perm p | overlay Sharpe | buy-hold Sharpe | exposure | turn/yr | placebo p |
|---|---|---|---|---|---|---|---|
| 1153 | -0.009 | 0.720 | 0.52 | 0.69 | 48% | 12.1 | 0.055 |
Signal last observed 2026-08-28.
B. CFTC E-mini S&P 500 non-commercial positioning
| n | corr(net/OI z, next-wk ret) | perm p | overlay Sharpe | buy-hold Sharpe | exposure | turn/yr | placebo p |
|---|---|---|---|---|---|---|---|
| 1181 | -0.017 | 0.521 | 0.55 | 0.69 | 51% | 5.1 | 0.029 |
Signal last observed 2026-08-28.
Reading
- Adoptable only if the overlay beats buy-hold on Sharpe AND the placebo p < 0.05 (random gate with the same exposure does not).
- A negative corr for B is the contrarian direction; a positive corr for A is the liquidity-tailwind direction. Sign alone is not a verdict.