On-line portfolio selection strategy with prediction in the presence of transaction costs (Q1397004)

The authors study a multi-period capital market with the aim of constructing a multiplicative on-line portfolio strategy which takes into account proportional transaction costs, and maximizes a functional \(F= \lambda F_w-F_T\), where \(F_w\) stands for the wealth increment, while \(F_T\) represents the transaction costs. They propose two different choices of \(F_w\) and \(F_T\) to derive two different update formulas for an on-line constructed portfolio making use of the Lagrange multiplier technique. The portfolio choice depends on the quality of stock prices' prediction, with a smaller \(\lambda\) indicating a smaller belief in the prediction accuracy, and vice versa. The proposed update formulas derived rigorously in Section 3, 4 and 5 seem to represent a valuable contribution to portfolio theory, based on a new method called the cross rate. The paper is not easy to read, although it is technically correct. It is proved that the strategy presented is superior with respect to the buy-and-hold strategy applicable to the best stock. Finally, this strategy has been tested on real data from LSE, demonstrating its advantage over a few well known methodologies, including that of Cover, and Helmbold.