Trading Strategy Model Based on Dynamic Programming

It is of great significance for traders to buy and sell volatile assets reasonably in market transactions. Traders will get different returns with different trading strategies. This paper studies the best trading strategies for different traders based on the daily prices of gold and bitcoin spanning a five-year period from September 11, 2016, to September 10, 2021. We propose a novel two-stage ARIMA model to forecast the prices of these assets based on historical data. The model achieves impressive correlation coefficients of 0.998 for gold and 0.999 for bitcoin, with RMSE values of 14.367 and 660.383, respectively. Based on these highly accurate predictions, we develop a trading strategy equation aimed at maximizing total assets, incorporating asset risk as a regularization term. The equation is solved with dynamic programming techniques. Furthermore, we evaluate the model’s generalization ability, robustness, yield, and sensitivity to transaction costs, resulting in three representative trading strategies: Insurance-type, Common-type, and Radical-type. Our findings demonstrate that the model exhibits strong resilience to market disturbances, provides returns capable of countering inflation, and aligns well with real-world cost sensitivities. These highlight the practical value of the proposed model in real-world trading scenarios.