A Novel Pairs Trading Model with Mean Reversion and Coefficient of Variance

N. Ekkarntrong, P. Sirisaengtaksin, P. Sattayatham, B. Premanode

Abstract

One of the problems of stock trading is that the stock prices are not consistent, depending on market condition. The Mean Reversion process of Pairs Trading is a market neutral strategy, which is independent of market movements and carried an assumption that any trading index will reverse to its mean value. This paper proposes a novel algorithm, called ‘multiclass Pairs Trading’, which is an advance of co-integration method in Pairs Trading technique. The proposed model uses Mean Reversion and coefficient of variance (CV) to analyze and classify a series of stocks to have different distribution. It provides a buffer-trading zone when the paired stocks are about to change their directions from high to low and vice versa. Moreover, this model extends an opportunity for any highly correlated and paired stocks to cross-trade with any lowly correlated and paired stocks. It serves to improve performance in portfolio trading. The 10-year data were collected from 127 stocks listed in the Global Dow. The results show that using the proposed model for the co-integrated Pairs Trading outperforms those of the conventional co-integrated Pairs Trading outstandingly. Thus, benefits of this model are not only mitigating stock trading risk but also maximise returns of them.