Quantitative Models and Factor Construction Quantitative Models and Construction Methods 1. **Model Name**: Dynamic Factor Adjustment Model **Model Construction Idea**: Combines factor momentum and reversal characteristics to dynamically adjust factor selection based on historical performance and failure probabilities[4][80][82] **Model Construction Process**: - Evaluate factor momentum using the average RankIC over the past 6 months and the average RankICIR over the past 3-12 months[4][82] - Calculate conditional failure probabilities by rolling one year of historical data to assess the likelihood of a factor transitioning from effective to ineffective[74][87] - Exclude factors with high failure probabilities and assign scores based on momentum and failure probabilities. Select the top N factors with the highest scores for equal-weighted scoring in each period[82][87][88] **Model Evaluation**: The model effectively balances momentum and reversal characteristics, reducing the impact of unstable factors and improving robustness in factor selection[82][87] 2. **Model Name**: "2+3" Dynamic Factor Model for Small-Cap Stocks **Model Construction Idea**: Combines two fixed factors (valuation and volatility) with three dynamically selected high-momentum factors to construct a robust small-cap stock selection model[98][99] **Model Construction Process**: - Fixed factors: Valuation (BTOP) and volatility (VOLATILITY) are always included due to their stable and significant performance in small-cap pools[98][99] - Dynamic factors: Exclude factors with conditional failure probabilities above 80% and select the top 3 factors based on medium- and long-term momentum scores[98][99] - Construct a portfolio of 50 equally weighted stocks based on the selected factors[98][103] **Model Evaluation**: The model demonstrates strong performance in small-cap pools, with high momentum and low reversal failure probabilities, making it robust against overfitting[98][103] 3. **Model Name**: "Exclusion + Scoring" Model for Large-Cap Stocks **Model Construction Idea**: Focuses on stricter exclusion of high-failure-probability factors and integrates failure information into the scoring process for large-cap stock selection[109][110] **Model Construction Process**: - Exclude factors with conditional failure probabilities above 70%[109][110] - Combine failure indicators into the momentum scoring model, selecting the top 5 factors with the highest comprehensive scores[109][110] - Construct a portfolio of 50 equally weighted stocks based on the selected factors[109][113] **Model Evaluation**: The model effectively addresses the high sensitivity and extreme reversals in large-cap pools, improving stability and performance[109][113] Model Backtesting Results 1. **Dynamic Factor Adjustment Model**: - Annualized return: 8.83% - Sharpe ratio: 0.42 - Excess annualized return: 11.47% - Maximum drawdown: 38.67%[103] 2. **"2+3" Dynamic Factor Model for Small-Cap Stocks**: - Annualized return: 8.83% - Sharpe ratio: 0.42 - Excess annualized return: 11.47% - Maximum drawdown: 38.67%[103] 3. **"Exclusion + Scoring" Model for Large-Cap Stocks**: - Annualized return: 8.40% - Sharpe ratio: 0.40 - Excess annualized return: 8.32% - Maximum drawdown: 36.40%[113] Quantitative Factors and Construction Methods 1. **Factor Name**: Valuation (BTOP) **Factor Construction Idea**: Measures the book-to-price ratio to capture undervalued stocks[8][39] **Factor Construction Process**: Calculate the ratio of book value to current market value for each stock[8][39] **Factor Evaluation**: Demonstrates stable and significant performance in small-cap pools, with strong selection ability in various market conditions[39][98] 2. **Factor Name**: Volatility (VOLATILITY) **Factor Construction Idea**: Measures the residual volatility of stock returns to identify low-risk stocks[8][50] **Factor Construction Process**: Calculate the standard deviation of residuals from a time-series regression of stock returns[8][50] **Factor Evaluation**: Performs well in both small-cap and large-cap pools, with low failure probabilities and consistent selection ability[50][98] 3. **Factor Name**: Earnings (EARNING) **Factor Construction Idea**: Measures earnings yield to capture profitability[8][39] **Factor Construction Process**: Calculate the ratio of earnings to market value for each stock[8][39] **Factor Evaluation**: Strong selection ability in large-cap pools, with stable performance across different market conditions[39][113] Factor Backtesting Results 1. **Valuation (BTOP)**: - RankICIR: Consistently ranks in the top 2 across small-cap pools[39][98] 2. **Volatility (VOLATILITY)**: - RankICIR: Demonstrates stable negative expression across all pools, with low failure probabilities[50][98] 3. **Earnings (EARNING)**: - RankICIR: Strong performance in large-cap pools, with high selection ability and stable expression[39][113]