Automated Formulaic Alpha Generation for Quantitative Investing using Evolutionary Algorithms

A cosmic ray consists of mostly highly energetic protons that emanate from the sun, the Milky Way and distant galaxies. By colliding with particles in our atmosphere they trigger a chain reaction that leads to so called cosmic-ray showers of lower energetic particles like pions [1]. In modern biology these are held responsible for inducing the random genetic mutations that led to the development of life on our planet as we know it [2]. In the frame of this thesis we will explore how we can make use of these random mutations of the genetic representation of competing candidates to find functions that correlate well with the stock market. This will yield a set of formulaic alphas that are used in quantitative investing to recognise patterns in a stock’s price development and trade on them accordingly. We will evaluate the performance of a set of eight formulaic alphas that are generated by a genetic program on the Nasdaq 100 and realize that they are highly correlated with the development of the federal reserve’s balance sheet. The assessment of a simple trading algorithm’s performance increase allows the assumption that the generated formulaic alphas help to recognise patterns in the stock market.