Pair trading with high-frequency data in the cryptocurrency market

Abstract

Purpose This study examines the performance of pair trading strategy in the cryptocurrency market under three statistical approaches including distance, cointegration and a hybrid method combining both distance and cointegration approaches. Design/methodology/approach The research uses daily, 4-h, 1-h, 15-min and 5-min data from the top 50 cryptocurrencies (by market capitalization) listed on Binance during three distinct periods: the bullish period of 2020, the stable period of 2021 and the bearish period of 2022. To perform a sensitivity analysis of the model, four approaches were implemented. First, both fixed and dynamic thresholds were applied across all three methods to assess their impact on trading results. Second, three standard deviations of 1.44, 1.65 and 2 were used, representing the coverage of normal data points in 85%, 90% and 95% of the time, respectively, to evaluate their influence on model profitability. Third, three different exit thresholds were employed to determine the extent to which changes in trade closure thresholds affect profitability. Fourth, the effect of the number of pairs in the portfolio on the model’s profitability was examined. Findings The findings from these approaches highlight the inefficiency of the cryptocurrency market and demonstrate the profitability of pair trading across various time frames, particularly in high-frequency time frames such as 15-min and 5-min intervals. Moreover, the results show that using a fixed threshold significantly outperforms a dynamic threshold in terms of both returns and Sharpe ratio. Additionally, the findings indicate a positive impact of altering the entry thresholds, exit thresholds and the number of pairs in the portfolio on the profitability of the models. Practical implications Due to the increasing attention to cryptocurrencies in investment management, the proposed model and the results of this study can be significantly used by cryptocurrency market traders, portfolio managers and fintech to achieve significant returns along with the increase in market liquidity. Originality/value This article has used pair trading strategies in the cryptocurrency market as a form of high-frequency trading for the first time. In addition, this article has proposed a hybrid approach based on the combination of distance and cointegration criteria to increase the efficiency of pair detection in the cryptocurrency market. It has evaluated the pair trading strategy in the form of different entry and exit criteria for a cryptocurrency.