InsidersTradesSigma
Quantitative Signals & Scoring
A statistical technique that transforms raw signal values into a standardized scale with mean zero and unit variance, enabling cross-sectional and temporal comparability of heterogeneous scoring inputs.
Z-standardization subtracts the mean from each observation and divides by the standard deviation, producing dimensionless scores centered at zero. In insider-trading and quant scoring platforms, this normalization is essential for combining signals with vastly different units, scales, and distributions, such as transaction dollar volume, execution price percentile, and conviction intensity. The resulting Z-scores facilitate fair weighting and aggregation across heterogeneous data sources, preventing large-magnitude but low-signal inputs from dominating composite scores.
In rolling-window implementations, Z-standardization is applied separately within each time period or cross-section to preserve temporal drift detection and identify regime shifts in insider behavior. This ensures that a score of +2.0 across different epochs or asset classes carries consistent statistical meaning, typically indicating the 97.7th percentile under normality. Practitioners must account for data snooping bias, survivorship effects in historical samples, and the sensitivity of standard deviation estimation to outliers when designing robust insider-scoring pipelines.
Formula