Factor Covariance and Shrinkage
The factor risk engine: how factor returns co-move, and why SurgeFlow shrinks the raw covariance matrix toward a structured target.
Category: Factor Research
Why covariance matters
The factor covariance matrix captures how factor returns move together. It is the input that translates a portfolio's factor exposures into portfolio risk. Two highly correlated factors carry overlapping risk, so true diversification is less than the number of factors suggests.
How this system builds it
SurgeFlow estimates a rolling covariance of the main factors over a 63-trading-day window, then applies Ledoit-Wolf linear shrinkage toward a structured target because raw sample covariance is noisy in short windows. The output stores the shrinkage intensity and a minimum-eigenvalue check so the matrix is verified positive semi-definite before any risk math uses it. - 63-day rolling window - Ledoit-Wolf shrinkage - Records shrinkage intensity - PSD / eigenvalue sanity check
How to read it
High off-diagonal correlation means two factors are nearly the same bet. The shrinkage-intensity field shows how much the raw estimate was regularized — higher intensity means a noisier sample that leaned more on the structured prior.