Lessons Learned Using Genetic Programming in a Stock Picking Context

Abstract

This is a narrative describing the implementation of agenetic programming technique for stock picking in aquantitatively driven, risk-controlled, US equityportfolio. It describes, in general, the problems thatthe authors faced in their portfolio context when usinggenetic programming techniques and in gainingacceptance of the technique by a skeptical audience. Wediscuss in some detail the construction of the fitnessfunction, the genetic programming system'sparametrisation (including data selection and internalfunction choice), and the interpretation andmodification of the generated programs for eventualimplementation.