Orient Securities·2025-07-20 05:44Quantitative Factors and Construction Methods 1. Factor Name: Beta - Construction Idea: Measures the sensitivity of a stock's return to market movements, capturing the market's preference for high Beta stocks [11] - Construction Process: Beta is calculated using Bayesian shrinkage to compress the market Beta [16] - Evaluation: Beta factor showed strong performance this week, with a return of 1.94%, indicating a sustained market preference for high Beta stocks [11][13] 2. Factor Name: Volatility - Construction Idea: Captures the market's preference for high-volatility assets [11] - Construction Process: Includes multiple metrics such as: - Stdvol: Standard deviation of daily returns over the past 243 days - Ivff: Idiosyncratic volatility from Fama-French 3-factor model over 243 days - Range: High/low price range over 243 days - MaxRet_6: Average return of the six highest-return days in the past 243 days - MinRet_6: Average return of the six lowest-return days in the past 243 days [16] - Evaluation: Volatility factor rebounded significantly this week, with a return of 0.82%, reflecting increased demand for high-volatility assets [11][13] 3. Factor Name: One-Year Momentum - Construction Idea: Measures the cumulative return over the past year, excluding the most recent month, to capture momentum effects [20] - Construction Process: Calculated as the cumulative return over the past 12 months, excluding the most recent month [20] - Evaluation: One-year momentum factor performed well in multiple indices, including: - CSI 500: Weekly return of 0.90% [27] - CSI 1000: Weekly return of 0.81% [35] - CSI All Share: Weekly return of 2.25% [47] 4. Factor Name: Standardized Unexpected Revenue (SUR) - Construction Idea: Measures the deviation of actual revenue from analyst expectations, standardized by the standard deviation of expected revenue [20] - Construction Process: $ SUR = \frac{\text{Actual Revenue} - \text{Expected Revenue}}{\text{Standard Deviation of Expected Revenue}} $ [20] - Evaluation: SUR factor showed strong performance across indices: - CSI 800: Weekly return of 1.37% [31] - CSI 1000: Weekly return of 0.86% [35] - CSI All Share: Weekly return of 1.53% [47] 5. Factor Name: Three-Month Reversal - Construction Idea: Captures short-term mean-reversion effects in stock prices [20] - Construction Process: Calculated as the cumulative return over the past three months, with a negative sign to reflect reversal [20] - Evaluation: Three-month reversal factor performed well in: - CSI 1000: Weekly return of 1.04% [35] - CNI 2000: Weekly return of 1.76% [39] --- Factor Backtesting Results 1. Beta Factor - Weekly Return: 1.94% - Monthly Return: 7.88% - Year-to-Date Return: 17.34% - Annualized Return (1 Year): 51.27% [13] 2. Volatility Factor - Weekly Return: 0.82% - Monthly Return: 1.86% - Year-to-Date Return: 5.96% - Annualized Return (1 Year): 27.16% [13] 3. One-Year Momentum Factor - CSI 500 Weekly Return: 0.90% [27] - CSI 1000 Weekly Return: 0.81% [35] - CSI All Share Weekly Return: 2.25% [47] 4. Standardized Unexpected Revenue Factor - CSI 800 Weekly Return: 1.37% [31] - CSI 1000 Weekly Return: 0.86% [35] - CSI All Share Weekly Return: 1.53% [47] 5. Three-Month Reversal Factor - CSI 1000 Weekly Return: 1.04% [35] - CNI 2000 Weekly Return: 1.76% [39] --- Factor Portfolio Construction: Maximized Factor Exposure (MFE) Construction Process - Objective Function: Maximize single-factor exposure $ \text{max } f^{T}w $ - Constraints: - Style exposure limits: $ s_{l} \leq X(w-w_{b}) \leq s_{h} $ - Industry exposure limits: $ h_{l} \leq H(w-w_{b}) \leq h_{h} $ - Stock weight deviation limits: $ w_{l} \leq w-w_{b} \leq w_{h} $ - Component stock weight limits: $ b_{l} \leq B_{b}w \leq b_{h} $ - No short-selling: $ 0 \leq w \leq l $ - Full investment: $ 1^{T}w = 1 $ - Turnover limits: $ \Sigma|w-w_{0}| \leq to_{h} $ [59][60][62] Backtesting Process 1. Set constraints for style, industry, and stock weight deviations 2. Construct MFE portfolios monthly 3. Calculate historical returns and risk metrics, adjusting for transaction costs [63][64]