| Baseline | ACTIVE | SUPER | TRIPLE | |
|---|---|---|---|---|
| 10d forward (recommended hold) | ||||
| Mean | — | — | — | — |
| Hit | — | — | — | — |
| 5d forward | ||||
| Mean | — | — | — | — |
| Hit | — | — | — | — |
| Signals | ||||
| Days / yr | — | — | — | |
| Total N | — | — | — | — |
| date | state | NL 5d | NFCI 5d | HYG−SPX |
|---|
Three tiers, each a superset of the next. ACTIVE when Fed liquidity is expanding and financial conditions are easing (~28/yr). SUPER adds credit confirmation — HYG outperforming SPX, a risk-on co-sign from the bond market (~6.5/yr). TRIPLE adds Primary Dealer Treasury inventory in the top tercile — dealer balance-sheet capacity confirming liquidity (~4.4/yr, highest conviction).
How to use it. An overlay, not an entry signal. Size up proportional to tier: ACTIVE = lean long, SUPER = size up, TRIPLE = high conviction. Hold for ~10 trading days (2 calendar weeks) — the Sharpe-optimal horizon. ACTIVE 10d: +164 bps / 86%. SUPER 10d: +220 bps / 86%. TRIPLE 10d: +240 bps / 89%. When nothing fires, don't flip short — draining regimes underperform but don't reliably lose.
Not for intraday. Underlying data (WALCL, NFCI) updates weekly. The 1-day edge is only +15 bps — real but too small after costs. Use GEX / order flow for intraday entries within the bias this filter sets.
Validation. Walk-forward with rolling 504-day point-in-time thresholds (no look-ahead). Out-of-sample (train 2015–23, frozen, test 2024–26): +108 bps / 79.5% at 5d, stronger than in-sample. Circular-permutation null test p<0.0001. Credit confirmation discovered via dedicated factor sweep (6 candidates tested; HYG−SPX divergence was the only one adding significant incremental lift).