Quantitative Models and Construction Methods - **Model Name**: Carhart Four-Factor Model **Construction Idea**: Incorporates momentum factor into the Fama-French three-factor model to capture the "stronger gets stronger" phenomenon in stock markets[14] **Construction Process**: Formula: $ R_{p}-r_{f}\!\sim\!\!\alpha+\beta_{1}(R_{M}-r_{f})+\beta_{2}(R_{M}-r_{f})^{2}+\beta_{3}S M B+\beta_{4}H M L+\varepsilon_{p} $ - $R_{p}-r_{f}$ represents excess return of the portfolio relative to the risk-free rate - $R_{M}-r_{f}$ represents market excess return - $SMB$ and $HML$ represent size and value premiums respectively[28][30] **Evaluation**: Widely applicable across various asset classes, but its effectiveness in predicting future returns in A-shares is limited due to strong short-term reversal effects[14][17] - **Model Name**: Industry-Stripped Alpha Momentum **Construction Idea**: Removes market and industry beta risks to isolate alpha returns for momentum factor construction[47] **Construction Process**: Formula: $ R_{p}-r_{f}{\sim}\alpha+\beta_{1}(R_{M}-r_{f})+\beta_{2}(R_{M}-r_{f})^{2}+\sum_{i=1}^{11}\beta_{2+i}\,l n d_{i}+\varepsilon_{p} $ - Adds industry index returns ($ln d_{i}$) to the regression model to strip industry beta risks[51] **Evaluation**: Improves stability compared to traditional momentum factors but shows weaker positive selection effects since 2019[52][53] - **Model Name**: Low-Diversification Momentum **Construction Idea**: Identifies dates with low fund diversification to reduce beta risk interference and enhance predictive power[5][56] **Construction Process**: - Groups fund daily returns by diversification levels (using standard deviation of returns) - Constructs three sub-factors: low-diversification return factor, sorting momentum factor, and Sharpe ratio factor - Combines these sub-factors equally to form the low-diversification momentum factor[65][93] **Evaluation**: Demonstrates strong predictive power with low correlation to traditional momentum factors, indicating reduced beta risk interference[93][104] Model Backtesting Results - **Carhart Four-Factor Model**: - Rank IC: 6.01% (past 122 days alpha momentum)[31] - Rank ICIR: 0.57 (past 122 days alpha momentum)[31] - Quarterly long-short win rate: 66.67%[31] - **Industry-Stripped Alpha Momentum**: - Rank IC: 7.81% (past 122 days)[53] - Rank ICIR: 0.97 (past 122 days)[53] - Quarterly long-short win rate: 69.92%[53] - **Low-Diversification Momentum**: - Rank IC: 10.10%[93] - Rank ICIR: 1.09[93] - Quarterly long-short win rate: 71%[93] - Annualized long-short return: 10.81%[98] Quantitative Factors and Construction Methods - **Factor Name**: Historical Return Factor **Construction Idea**: Uses past fund returns to predict future performance[19] **Construction Process**: - Calculates returns over different time windows (e.g., past 20, 61, 122 days) - Tests predictive power using Rank IC and Rank ICIR metrics[20][22] **Evaluation**: Short-term returns show weak predictive power; long-term returns improve prediction but remain unstable[22][23] - **Factor Name**: Sharpe Ratio Factor **Construction Idea**: Adjusts fund returns for volatility to improve stability[24] **Construction Process**: - Calculates Sharpe ratios over different time windows (e.g., past 20, 61, 122 days) - Tests predictive power using Rank IC and Rank ICIR metrics[25][26] **Evaluation**: Stability improves compared to historical return factor but fails to address beta risk interference effectively[26][27] - **Factor Name**: Low-Diversification Return Factor **Construction Idea**: Focuses on low-diversification dates to reduce beta risk interference[65] **Construction Process**: - Groups fund daily returns by diversification levels - Uses average returns of the lowest-diversification group as the factor score[65][67] **Evaluation**: Strong predictive power with stable performance across different time windows[67][72] Factor Backtesting Results - **Historical Return Factor**: - Rank IC: 6.44% (past 244 days)[20] - Rank ICIR: 0.54 (past 244 days)[20] - Quarterly long-short win rate: 59.35%[20] - **Sharpe Ratio Factor**: - Rank IC: 6.44% (past 244 days)[25] - Rank ICIR: 0.64 (past 244 days)[25] - Quarterly long-short win rate: 61.79%[25] - **Low-Diversification Return Factor**: - Rank IC: 10.03% (past 3 months)[68] - Rank ICIR: 1.06 (past 3 months)[68] - Quarterly long-short win rate: 69.11%[68]