ABSTRACT
While conventional equal-risk approaches provide a solid foundation, investors can further improve on the efficiency of their portfolios. This chapter describes advanced portfolio construction techniques that are relevant for both hedge fund investors and managers. It covers several machine learning approaches, two recent cutting-edge methods, and several interesting complementary nuggets. Equally volatility-adjusted completely ignores any potential links among assets, potentially leading to concentrated portfolios. The chapter illustrates this issue by considering a simple example with four assets A, B, C and D, where the assets are mostly uncorrelated except assets A and B have a pair-wise correlation of 0.9. The tree structure helps visualize a top-down hierarchy that is relevant for institutional investors. The HRP approach of Marcos Lopez de Prado applied graph theory and machine learning techniques to build well-diversified portfolios that performed well out-of-sample. The methodology relies on the top-down tree structure and follows three steps: tree clustering, quasi-diagonalization, and recursive bisection.
