Genetic Programming for Equity Factor Discovery and Testing
Summary
The document explains how genetic programming can evolve formulas from stock price and volume inputs to discover candidate equity selection factors. It outlines the evolutionary cycle of selection, crossover, and mutations, and describes adapting a genetic programming toolkit with time-series functions, factor-testing routines, neutralization, and parallel computation. Candidate formulas are evaluated using average Rank IC against a forward return target, followed by regression, information coefficient, portfolio-sorting, decay, and correlation analyses.
In a historical test on Chinese A-shares, the researchers report six candidate factors with some incremental predictive information after controlling for industry, size, and several trading-style exposures. Their behavior was mixed: some factors were more stable, while others showed weaker or nonlinear results, and two related factors were more correlated. The study warns that factors may fail, become hard to interpret, and may not generalize beyond the tested stock universe. Its findings are historical evidence, not assurance of future performance.
Key ideas
- Genetic programming evolves formula trees to search for candidate factors from price and volume data.
- Fitness can be based on the mean cross-sectional Rank IC with forward returns.
- Factor evaluation can combine neutralization, regression, IC analysis, portfolio sorting, and decay checks.
- The reported Chinese equity study found six candidates with mixed stability and correlation.
- Historically discovered factors can fail and may not transfer to other universes.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.