Description:
The framework for mean-variance optimization in finance has been discussed since the 1950s. Recently, to improve computation efficiency on a large number of assets, an optimal dynamic portfolio model was introduced to speed up the search for the trajectory of allocated weights, which used the prefactor of γ/2 for risk in the dynamic setting of t . This efficient model can be further transformed into a new binary unconstrained model, which is preferred by the new generation of quantum annealing solvers. In this study, we analyze the US stock and cryptocurrency markets employing the optimal dynamic portfolio with Gurobi solver and binary unconstrained version of the model with quantum annealing solver. Then we compare the out-of-sample performance of the models to the corresponding ETF, benchmark, and 1/N naïve diversification model. As a benchmark for the cryptocurrency markets, we use Bitcoin mining companies as the benchmark: Riot Blockchain Inc. and Marathon Digital Holdings Inc. We also analyze how the COVID-19 pandemic, market turmoil, and high-frequency data affect the portfolio selection and its performance.The results show that both ODP and BUODP models can find the solution very quickly, even with the S&P500 market within half a minute, and the out-of-sample performance of the BUODP model is superior to the ODP model and competitive with the corresponding ETF. We also find that using the data from DJIA stocks and cryptocurrencies, the mean-variance optimization models can further mitigate the risk and perform better than the ETF and 1/N naïve diversification model for the second half of 2021