Pairs trading in cryptocurrency markets: A comparative study of statistical methods

Abstract

This study aims to identify a secure and efficient trading approach for investors in highly volatile cryptocurrency markets. While pairs trading is a promising strategy, the available literature on this topic in cryptocurrency markets is limited and primarily based on traditional methods. This study compares six statistical approaches for selecting trading pairs: Cointegration, Correlation, Distance, Fluctuation Behaviour, Hurst Exponent, and Stochastic Differential Residual. We utilise intraday data from 30 cryptocurrencies on the Binance exchange during a three-month period from 1 January 2022 to 31 March 2022. The trading results are measured by 11 criteria covering return, risk, and trade aspects. The findings show that Distance performs well at all three frequencies of 1 minute, 5 minutes, and 60 minutes, with a total return of 208.12%, 236.31%, and 210.36%, respectively. At the 60-minute frequency, this method exhibits low risk and performs impressively even when all other techniques generate negative returns, as well as displays the lowest number of expired counts and the highest success ratio among the methods. At 1-minute and 5-minute frequencies, Cointegration and Hurst Exponent produce results comparable to those of Distance. The Student t-test conducted confirms these conclusions.