A reality check on trading rule performance in the cryptocurrency market: Machine learning vs. technical analysis

This paper performs a reality check for the superior predictive ability of Machine Learning and Technical Analysis trading rules in the cryptocurrency market. After controlling for data snooping and various market frictions, we find that statistically significant positive excess returns are rarely achieved, independent of the data sampling frequency, type of trading position, or test significance level. Also, cross-sectional performance is correlated with risk factors such as beta and idiosyncratic volatility, implying that trading rules mostly capture market risk premiums. Overall, trading rules do not seem to provide additional benefits in cryptocurrency markets compared to traditional financial markets.