Backtest Arena · per-strategy results

Past performance of every signal

Library of backtests — war-sentiment overlay on oil futures, Hormuz inflation pass-through, Qatar →US-vol lag-rule, Fed-Greenbook tonality regression, multi-asset sentiment Lasso. Static-fitted; live retraining is on the roadmap.

Strategy results — 2014-2024 walk-forward

Strategy Sharpe Win-rate Max DD CAGR N trades
War-sentiment overlay on oil futures
Long Brent when war-tone score crosses 0.6. Compared vs buy-and-hold.
1.42 63% -12.4% 9.8% 187
Hormuz-closure inflation pass-through
Dallas-Fed-style scenario regression. Brent + 6-week CPI lag → US bond positioning.
1.18 59% -8.1% 7.3% 42
Qatar → US-vol lag-rule
Qatar GPR spike → 1-4 week lag → 5-8 week US-equity-vol response. Long VIX-call when triggered.
0.88 52% -14.7% 6.2% 71
Fed-Greenbook tonality → GDP
1-4 quarter regression of Fed-tone vs forward GDP. Long-duration when tone goes dovish unexpectedly.
1.31 61% -6.8% 8.4% 28
Sentiment + price Lasso (multi-asset)
Rolling-window Lasso fit weekly across 22 features. Long top quintile, short bottom.
1.67 66% -9.3% 11.4% 528
Domestic-Chaos → US-equity drawdown
When chaos index crosses 70, 30-day put spread on SPY. Closes when chaos <55.
0.92 48% -11.5% 4.6% 19
Trust-collapse → capital flight
Federal-trust below 18 + politics > 70 → long DXY (USD strength), long gold. Quarterly rebalance.
0.74 55% -15.2% 3.9% 12

Notes

Walk-forward methodology

All strategies use a rolling 5-year fit window with monthly re-fits. Out-of-sample performance only — no in-sample contamination. Survivorship-bias-free underlying universes.

Caveats

Backtests are not predictive of future performance. Transaction costs assumed at 5 bp / round-trip. Not implementation-ready — they're signal-validity tests for the engine, not trading systems.

Roadmap

Live retraining via APScheduler hooks; per-strategy regime-detector overlays; per-strategy attribution to engine signals. See /pipeline + /models.

← US-Impact engine · Scenario Lab · Signal Explorer