金工定期报告20250806：优加换手率UTR2.0选股因子绩效月报-20250806

Quantitative Factors and Construction Methods 1. Factor Name: UTR (U-Turnover Rate) - Construction Idea: The UTR factor combines two sub-factors, "Turn20" (volume-small factor) and "STR" (volume-stable factor), using a scoring method to address the issue of "1+1<2" in factor integration. The key idea is to prioritize stocks with stable volumes while favoring higher turnover within this stable group [6] - Construction Process: 1. At the end of each month, calculate the "Turn20" and "STR" values for all stocks [6] 2. Rank all stocks by the "STR" factor in ascending order and assign scores from 1 to N (N is the total number of stocks), referred to as "Score 1" [6] 3. For the top 50% of stocks ranked by "STR", rank them by "Turn20" in descending order and assign scores from 1 to N/2, referred to as "Score 2". The final score for these stocks is "Score 1 + Score 2" [6] 4. For the bottom 50% of stocks ranked by "STR", rank them by "Turn20" in ascending order and assign scores from 1 to N/2, referred to as "Score 3". The final score for these stocks is "Score 1 + Score 3" [6] 5. The resulting factor is named "UTR" [6] - Evaluation: The UTR factor effectively integrates the two sub-factors, achieving the goal of favoring higher turnover within stable-volume stocks [6] 2. Factor Name: UTR2.0 (U-Turnover Rate 2.0) - Construction Idea: UTR2.0 improves upon the original UTR factor by transitioning from ordinal scale to ratio scale for factor values. Additionally, it introduces a coefficient for the "Turn20" factor, which is a function of "STR", to better capture the varying impact of "Turn20" across different stability levels. The softsign activation function from neural networks is used to model this relationship [7] - Construction Process: 1. Transition from ordinal scale to ratio scale for factor values to retain more information [7] 2. Define the coefficient for "Turn20" as a function of "STR", where the coefficient increases with stability (positive impact) and decreases with instability (negative impact) [7] 3. Use the softsign activation function to model the relationship: $\mathrm{UTR2.0} = \mathrm{STR} + \text{softsign}(\mathrm{STR}) \cdot \mathrm{Turn20}$ $\text{softsign}(x) = \frac{x}{1 + |x|}$ [7] - Evaluation: Compared to the original UTR factor, UTR2.0 achieves lower returns but demonstrates superior volatility, information ratio (IR), and monthly win rate, making it a more robust factor [7][9] --- Factor Backtesting Results 1. UTR2.0 Factor - Annualized Return: 40.36% [9] - Annualized Volatility: 14.97% [9] - Information Ratio (IR): 2.70 [9] - Monthly Win Rate: 75.74% [9] - Maximum Drawdown: 11.03% [9] 2. July 2025 Performance (UTR2.0 Factor) - Long Portfolio Return: 1.29% [9] - Short Portfolio Return: -0.06% [9] - Long-Short Portfolio Return: 1.35% [9]