Heterotic Risk Models for Equity Portfolio Optimization

Summary

The document describes a method for building heterotic risk models for equities. The approach combines industry classification with principal-component covariance estimates within stock subgroups, then reduces the covariance model through a recursive, Russian-doll construction. The authors present the resulting models as a way to build stable risk estimates from short lookback periods and say they provide a complete algorithm and source code.

The stated application is Sharpe ratio optimization for intraday mean-reversion signals based on overnight returns. The authors report that optimization with heterotic models improves performance characteristics compared with weighted regressions using principal components or industry classifications. They also describe source code for statistical risk model construction and optimization under homogeneous linear constraints and position bounds. The supplied text does not specify the data, evaluation period, metrics, or magnitude of the reported improvement, so the result cannot be independently assessed from this summary alone.

Key ideas

• The method combines industry groupings with principal-component covariance estimates within stock subgroups.
• A Russian-doll construction reduces the size of the factor covariance matrix.
• The authors present the approach as suitable for stable risk estimates using short lookback periods.
• They report improved optimization characteristics for intraday mean-reversion signals based on overnight returns.
• The described optimization includes homogeneous linear constraints and position bounds.

Tags

# Heterotic Risk Models


# Heterotic Risk Models









We give a complete algorithm and source code for constructing what we refer to as heterotic risk models (for equities), which combine: i) granularity of an industry classification; ii) diagonality of the principal component factor covariance matrix for any sub-cluster of stocks; and iii) dramatic reduction of the factor covariance matrix size in the Russian-doll risk model construction. This appears to prove a powerful approach for constructing out-of-sample stable short-lookback risk models. Thus, for intraday mean-reversion alphas based on overnight returns, Sharpe ratio optimization using our heterotic risk models sizably improves the performance characteristics compared to weighted regressions based on principal components or industry classification. We also give source code for: a) building statistical risk models; and ii) Sharpe ratio optimization with homogeneous linear constraints and position bounds.

Shown in full with attribution under the source's licence. Licence: abstract CC0

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.