PAMR: Passive aggressive mean reversion strategy for portfolio selection

Abstract

This article proposes a novel online portfolio selection strategy named "Passive Aggressive Mean Reversion" (PAMR). Unlike traditional trend following approaches, the proposed approach relies upon the mean reversion relation of financial markets. Equipped with online passive aggressive learning technique from machine learning, the proposed portfolio selection strategy can effectively exploit the mean reversion property of markets. By analyzing PAMR's update scheme, we find that it nicely trades off between portfolio return and volatility risk and reflects the mean reversion trading principle. We also present several variants of PAMR algorithm, including a mixture algorithm which mixes PAMR and other strategies. We conduct extensive numerical experiments to evaluate the empirical performance of the proposed algorithms on various real datasets. The encouraging results show that in most cases the proposed PAMR strategy outperforms all benchmarks and almost all state-of-the-art portfolio selection strategies under various performance metrics. In addition to its superior performance, the proposed PAMR runs extremely fast and thus is very suitable for real-life online trading applications. The experimental testbed including source codes and data sets is available at http://www.cais.ntu.edu.sg/~chhoi/PAMR/.

Moving average reversion strategy for on-line portfolio selection

“…Although some recently proposed mean reversion algorithms [46,45] have achieved promising results on many real datasets, they might perform extremely poorly on certain datasets, such as the DJIA dataset [8]. Comparing with Borodin et al [8], which exploits multi-period correlation, we found that the assumption of single-period prediction may attribute to the performance degradation.…”

“…Li et al [46] proposed Passive Aggressive Mean Reversion (PAMR) strategy, which exploits the mean reversion property with the Passive Aggressive (PA) online learning [53,17]. The main idea of PAMR is to design a loss function of (b; x t ) in order to reflect the mean reversion property, that is, if the expected return based on last price relative is larger than a pre-defined threshold, the loss will linearly increase; otherwise, the loss is zero.…”

“…show that OLMAR not only can achieve better performance (in terms of cumulative wealth) than existing algorithms, and but also can avoid the significant performance degradation on certain datasets, such as the DJIA dataset [8,46].…”

“…Comparing with Borodin et al [8], which exploits multi-period correlation, we found that the assumption of single-period prediction may attribute to the performance degradation. On the other hand, as illustrated in existing studies [46,45], Borodin et al [8] cannot fully exploit the potential of (multi-period) mean reversion for its heuristical nature.…”

“…One important property exploited by many existing studies [8,46,45] is the mean reversion property, which assumes poor performing stocks will perform well in the subsequent periods and vice versa. Although some recently proposed mean reversion algorithms [46,45] have achieved promising results on many real datasets, they might perform extremely poorly on certain datasets, such as the DJIA dataset [8].…”