Discovery Engine Results
We tested 15 hypotheses across volatility, macro, crypto, FX, and sector data. After Bonferroni correction for multiple testing, 3 relationships remain significant.
Methodology: Each hypothesis was tested using permutation testing (10,000 iterations) for statistical significance, with 70/30 train/test splits to ensure generalization. We applied Bonferroni correction to account for multiple testing (α = 0.05/n). A "hit" is a 2%+ move in the target asset within 5 trading days of the signal.
When VIX spikes 10%+ in a single day, SPY experiences significant volatility within 5 trading days.
VIX spikes signal incoming equity turbulence. A 50% hit rate vs 25% random baseline means this signal doubles your odds of catching a 2%+ move.
When high-yield bonds (HYG) drop 0.5%+ in a day, equities follow with significant moves.
Credit markets lead equities. When HYG stress appears, SPY moves 2%+ within 5 days 81% of the time vs 22% on random days. The signal generalizes perfectly.
When news coverage spikes in oil-sensitive regions, crude oil moves significantly within 5 trading days.
Geopolitical events in oil-sensitive regions (Middle East, Russia, Venezuela, Libya) precede significant oil price moves. This isn't prediction — it's pattern recognition.
These aren't predictions — they're patterns. We systematically test relationships between data sources and market movements. When we find statistically significant correlations that generalize out-of-sample, we surface them.
More discoveries coming as we expand our data coverage.