A Multi-Factor Quantitative Study of A-Share Stocks Following the Daily Price Limit Using Machine Learning

Abstract

This study examines A-share Main Board stocks that triggered the daily price limit between 2023 and 2025, comparing the performance of a conventional linear multi-factor framework against a random forest algorithm in short-term trading. Empirical results demonstrate that traditional rule-based screening produces a significantly negative expected return, whereas the random forest model effectively captures nonlinear interactions among micro-level price-volume factors and consecutive-limit-streak features. Under rigorous out-of-sample backtesting, the machine-learning strategy achieves a qualitative improvement in both win rate and Sharpe ratio while substantially reducing maximum drawdown. The present study contributes empirical evidence on microstructural pricing anomalies in the A-share market and provides a high-win-rate reference framework for quantitative short-term trading.