Aggregating reverse expert strategies considering extreme market situations for online portfolio selection

Online portfolio selection is a decision-making process that involves dynamically adjusting asset positions based on historical price sequences. In the decision-making process, extreme market situations have a significant impact. Due to investors’ overreaction, asset prices often reverse after experiencing extreme market situations. However, existing online portfolio strategies seldom consider the significant impact of extreme market situations on investment decisions. In this paper, we construct a novel online portfolio strategy based on reversal signal and online gradient update algorithm. First, we identify reversal signal through a moving window, predict asset prices via the reversal signal, and construct a series of expert strategies based on predicted value of asset prices. Second, we aggregate the expert strategies through the online gradient update algorithm and propose our strategy. Then, we theoretically prove that the regret of our strategy has an upper bound, which guarantees the competitive performance of our strategy. Finally, we conduct extensive numerical experiments using real financial data from different markets. The results show that the proposed strategy performs well on cumulative wealth, risk-adjusted returns and transaction costs.