Risk Forecast Methodology
Definitions for the published factor-validation table and how these diagnostics differ from portfolio risk-forecast research.
These rows are factor-return diagnostics. The separate portfolio-volatility study compares predicted and realized volatility across portfolios and has its own sample, model version, uncertainty and target. Passing the table thresholds below does not establish that the portfolio study has met its target. See the Factor-model research evidence.
Latest validation window
As of Oct 22, 2026. Computed 13.5d ago.
Validation metrics
Information Coefficient
The table producer uses each factor's previous observed return as its prediction, computes daily Spearman correlations across eligible factors, then averages those daily correlations over the validation lookback.
-0.037
Target > 0.50
Daily rank agreement between lagged and realized factor returns across factors.
Bias Ratio
Values near 1.00 mean forecasted risk and realized risk are calibrated.
1.002
Target 0.98-1.02
Ratio of predicted volatility to realized volatility across the validation sample.
95% VaR Coverage
Computed from realized returns versus the predicted two-sided 95% risk interval.
92.2%
Target 93.0%-97.0%
Share of realized returns contained inside the model's predicted 95% interval.
Per-factor performance
Quality score
Each factor is scored from 0 to 100 by multiplying positive IC strength, 12-month half-life IC stability, IC coverage, and the latest outlier penalty.
Outlier and variance checks
The validation dashboard flags factor-return moves above five rolling standard deviations, shows 90-day factor-return autocorrelation, and reports each factor contribution to rolling factor variance.