Adaptive Factor Trading via Market Regime Clustering Evidence from the CSI 300 Index

AbstractTraditional multi-factor strategies often employ a fixed trading signal threshold irrespective of market conditions. This paper proposes an adaptive thresholding framework that first clusters daily market environments using K-means on 13 technical indicators, then optimises a composite factor signal's entry/exit threshold separately for each regime. Applying a strict walk-forward test on CSI 300 data from 2010–2025, we find that the regime-adaptive strategy delivers an annualised return of 10.68% and a Sharpe ratio of 0.81, substantially outperforming a globally optimised threshold (8.68%, 0.72), a conventional fixed threshold (7.12%, 0.53), and the buy-and-hold benchmark (1.63%, 0.18). The gains stem from improved risk control rather than from market timing. The method offers a simple yet effective enhancement for systematic multi-factor investing. KeywordsAdaptive threshold, CSI 300 index, Factor investing, K-means clustering, Market regimes, Technical indicators, Walk-forward validation.