Clustering-augmented reversal strategy improves return performance: Evidence from Chinese stock market

This paper proposes a novel clustering-augmented reversal strategy that employs the unsupervised K-means algorithm to group stocks before constructing contrarian portfolios. This strategy consistently outperforms the standard reversal approach in the Chinese stock market, generating significant monthly long-short returns. The performance gains are primarily driven by short positions, while long positions remain insignificant. Notably, technical characteristics prove more effective than fundamental characteristics in clustering stocks, largely due to the retail-dominated structure of the A-share market. Clustering proves crucial, contributing approximately 20 %–45 % of the total portfolio returns. Analysis of return quantile transition distributions further confirms that reversals are more pronounced among past winners than among losers. Finally, risk-adjusted regressions reveal that the long-short portfolios generate significantly positive alphas while exhibiting no significant loadings on risk factors, underscoring a risk-offsetting mechanism within clusters.