Every candidate energy-market trading signal goes through the same five-step validation pipeline before it's classified: predictive content, monotonicity, regime stability, mandatory walk-forward out-of-sample testing, and economic significance measured against realized volatility. Most of what we test doesn't survive that bar - we publish the tiering honestly, not just the winners.
Natural gas storage level vs. its own 5-year seasonal average. The single strongest, most cross-family-robust predictor in the whole library - confirmed 4 separate times.
WTI/Brent relative-value spread. A clean sweep - 100% on every scorecard dimension, 27 years of walk-forward folds.
ERCOT settlement price vs. Henry Hub gas cost, all 15 real hubs/load zones. 100% on every scorecard dimension - the newest Tier 1 entry.
Regional heating-degree-day anomalies vs. forward Henry Hub returns. Real, regime-dependent, and independently useful alongside the Tier 1 factors.
CAISO and NYISO spark spread are real, already-computed factors - held back from the public API for legal and backtestable-history reasons, not shown here. Full breakdown of every tier, including the excluded subjects within otherwise-public factors (a real region can be excluded even when its own factor family is public): FACTOR_CATALOGUE.md.
Full validation methodology and every real number behind these verdicts: FACTOR_CATALOGUE.md.
Free early access to the API: Get early access →