Pair trading strategies in the cryptoassets market: a cointegration framework with optimized thresholds using genetic algorithms

The crypto-assets markets are notoriously volatile and risky. In this context, market-neutral type strategies, such as pair-trading, may be relevant. In this paper, we focus on the implementation of pair-trading strategies with a wide range of crypto-assets over periods between August 2021 and January 2024. To carry out this study, we combine econometric and machine learning techniques which differ from those used in existing literature on the subject. By using cointegration tests and error correction models, we identify a sample of 229 pairs suitable for pair-trading strategies. Using a genetic algorithm and pair clustering, we test four strategies using standard and optimized thresholds. The results highlight the existence of profitable cointegrating relationships, and, therefore, short-term market inefficiencies in the crypto-assets market. Indeed, though still risky, the best strategy identified in terms of risk-return trade-off, with a median maxdrawdown of 15.29%, delivers an average annual Sharpe ratio per pair of 0.69 over the out-of-sample period.