InsidersTradesSigma
Quantitative Signals & Scoring
A statistical measure of the correlation between a signal's predictions and actual subsequent asset returns, normalized to account for forecast accuracy and used to validate the predictive power of insider trading indicators and quant scoring models.
The Information Coefficient (IC) quantifies the rank correlation between forecasted signals (such as insider transaction propensity scores or behavioral anomaly flags) and realized forward returns over a specified holding period. In quant insider-trading platforms, IC serves as a critical backtesting metric, revealing whether the signal genuinely contains alpha or merely correlates with market noise. A positive IC near +1 indicates strong predictive alignment, while values near 0 suggest the signal lacks discriminatory power. The IC is typically computed using Spearman's rank correlation to remain robust against outliers and non-linear relationships inherent in insider activity patterns.
In compliance and surveillance workflows, IC analysis helps distinguish legitimate alpha-generating signals from spurious correlations that might trigger false positives in market abuse detection. By decomposing IC into individual signal components (insider filing velocity, closely-associated-person patterns, Form 144 timing), teams isolate which behavioral indicators truly predict price movement versus which merely reflect post-trade market microstructure. Degradation of IC over time (signal-decay) necessitates periodic recalibration of scoring thresholds and feature weights to maintain detection efficacy and reduce type-II errors in insider-trading identification.
Formula