国君研究|主动量化观点 · 合集

Quantitative Models and Construction Methods 1. Model Name: Style Timing Framework - Model Construction Idea: The model aims to capture the cyclical nature of different asset classes within the A-share market, particularly focusing on high dividend assets, micro-cap stocks, and mid-cap assets[3]. - Model Construction Process: - High Dividend Assets: The model identifies periods of significant excess returns in January-April, August, and November-December, while noting significant pullbacks in May-June[3]. - Micro-cap Stocks: The model considers regulatory changes and delisting systems as key factors. It identifies that a week of stable rebound often signals a short-term bottom[4]. - Mid-cap Assets: The model uses the CSI 1000 index as a representative of mid-cap assets. It identifies cyclical fluctuations in average investment returns, with the index's starting point often at an average return below -5% and a high point at a return above 10%[5]. - Model Evaluation: The model effectively captures the cyclical nature of different asset classes, providing insights into timing strategies for high dividend assets, micro-cap stocks, and mid-cap assets[3][4][5]. 2. Model Name: Industry Comparison Framework - Model Construction Idea: The model supplements traditional industry rotation strategies by incorporating the perspective of stock price pressure, assuming that incremental funds are evenly allocated across sectors[6]. - Model Construction Process: - The model uses the free float distribution model to quantify changes in shareholder behavior. - It generates the final industry selection indicators by equally weighting two factors. - Factor 1: Long-only portfolio annualized return of 2.2%, exhibiting strong style characteristics but with poor stability. - Factor 2: Long-only portfolio annualized return of 11%, achieving approximately 13% excess return over the CSI 300 index, with a long-short portfolio return of 18%, but with significant drawdowns in excess returns. - Composite Factor: Combining the two factors equally, the final factor significantly improves performance, with the long-only portfolio annualized return increasing to around 13%, relative excess return of 16%, and the short-only portfolio return of -10%, with drawdowns in excess returns reduced to around 10%[6]. - Model Evaluation: The composite factor significantly enhances the performance of the industry selection strategy, providing robust excess returns with reduced drawdowns[6]. Model Backtesting Results Style Timing Framework - High Dividend Assets: Significant excess returns in January-April, August, and November-December; notable pullbacks in May-June[3]. - Micro-cap Stocks: A week of stable rebound often signals a short-term bottom[4]. - Mid-cap Assets: Average investment returns below -5% indicate a starting point, while returns above 10% indicate a high point[5]. Industry Comparison Framework - Factor 1: Long-only portfolio annualized return of 2.2%[6]. - Factor 2: Long-only portfolio annualized return of 11%, excess return of 13% over the CSI 300 index, long-short portfolio return of 18%[6]. - Composite Factor: Long-only portfolio annualized return of around 13%, relative excess return of 16%, short-only portfolio return of -10%, drawdowns in excess returns reduced to around 10%[6].