Browsing by Author "Bazaliy, Viacheslav"

This thesis provides an analysis of Growth Optimal Portfolio (GOP) in discrete time. Growth Optimal Portfolio is a portfolio optimization method that aims to maximize expected long-term growth. One of the main properties of GOP is that, as time horizon increases, it outperforms all other trading strategies almost surely. Therefore, when compared with the other common methods of portfolio construction, GOP performs well in the long-term but might provide riskier allocations in the short-term. The first half of the thesis considers GOP from a theoretical perspective. Connections to the other concepts (numeraire portfolio, arbitrage freedom) are examined and derivations of optimal properties are given. Several examples where GOP has explicit solutions are provided and sufficiency and necessity conditions for growth optimality are derived. Yet, the main focus of this thesis is on the practical aspects of GOP construction. The iterative algorithm for finding GOP weights in the case of independently log-normally distributed growth rates of underlying assets is proposed. Following that, the algorithm is extended to the case with non-diagonal covariance structure and the case with the presence of a risk-free asset on the market. Finally, it is shown how GOP can be implemented as a trading strategy on the market when underlying assets are modelled by ARMA or VAR models. The simulations with assets from the real market are provided for the time period 2014-2019. Overall, a practical step-by-step procedure for constructing GOP strategies with data from the real market is developed. Given the simplicity of the procedure and appealing properties of GOP, it can be used in practice as well as other common models such as Markowitz or Black-Litterman model for constructing portfolios.