Bonds
US Treasury Yield-Curve Spreads
us treasury yield-curve spreads — regime-aware mean reversion across the long end.
Trades one thing well: the tendency of two linked US Treasury futures to snap back to their normal spread — a single mean-reversion specialist; the other engines are deliberately off (they'd trade the wrong way on a stationary spread).
source · S2 §1, §3, §10The link can break; a health check blocks trading when it does but cannot predict a regime change; hedge ratios are approximations; financing/venue risk remains.
source · S2 §1, §3, §10Calibration criteria are gates, not outcomes; no realized-performance claim is made.
engines & regime · methodology
A single mean-reversion specialist on a DV01-weighted spread residual — the other canonical engines are deliberately OFF, since trend/breakout logic would trade the wrong way on a stationary spread.
Two leg prices → DV01-weighted residual → rolling z-score → entry when the z-score is stretched. A cointegration-health check measures rolling-mean drift in sigma units and blocks entries when the relationship degrades. The hybrid-AI overlay gates but is a no-op on the live spread path.
Calibration criteria are applied as gates, not outcomes — a configuration either clears the gate or it does not ship.
source · S2 §2-4, §9The cointegration that grounds the edge is a statistical relationship that can and does break; the health check detects degradation but cannot anticipate a structural regime change in curve dynamics, and the architecture treats a sustained drift breach as a block, not a tradable signal.
source · S2 §10- ✓available
Whitepaper
S2 RX07 SPRD_MR Bonds 5-leg paper exists in OneDrive 17 Bond/FX Algo Upgrades
- ○pending
Standing requirements
- ✓available
Live heatmap
S2 RX07 100%-parameter heatmap shipped — 16/16 SHIP, leak-free (40fa41f)
- ◐partial
Calibration matrix
Q7_Calibration_Matrix_S2.xlsx exists locally; UI render pending
- ◐partial
Change / defect / repair log