Gamma and Vega Hedging Using Deep Distributional Reinforcement Learning

John Hull, Senior Fellow, GRI and Professor, Joseph L. Rotman, School of Management, University of Toronto

Jay Cao, Jacky Chen, Soroush Farghadani, Zissis Poulos, Zeyu Wang, and Jun Yuan Joseph L. Rotman, School of Management, University of Toronto

Several pieces of Canadian money in a pile

Overview

Machine learning is a tool being used for developing trading hedging strategies. This study investigates the use of deep reinforcement learning algorithms to hedge convexity and volatility (gamma and vega) in a system that includes transaction costs. It examines the optimal hedge strategy under three different objective functions: mean-variance trade-off, value at risk, and conditional value at risk.