Discovery of multi-spread portfolio strategies for weakly-cointegrated instruments using boosting-based optimization

Abstract

Increasing complexity of the modern market dynamics requires new quantitative frameworks for the discovery of stable portfolio strategies. Important requirements include the ability of the coupled and self-consistent optimization of the dynamic strategies and asset allocations as well as robust built-in mechanisms for the strategy complexity control to ensure acceptable out-of-sample performance. Recently introduced boosting-based optimization naturally incorporates all these features. Originally, the framework was described as a generic tool for the discovery of compact portfolio strategies from a given pool of existing financial instruments and base trading strategies. Here I outline the important generalization of this framework that allows simultaneous discovery of new synthetic instruments represented as generalized spreads of existing financial instruments and dynamic trading strategies for each such spread. Detailed arguments and real-market example clarify the essence of this new framework as a powerful generalization of the exiting pairs trading strategies and cointegration-based techniques.