收益差择时模型：基于A股指数与恒生指数的实证

Quantitative Models and Construction Simple Return Model - Model Name: Simple Return Model - Construction Idea: The model uses the simple return of closing prices to track trends and make trading decisions [12][13] - Construction Process: 1. Calculate the simple return as: $ \text{Simple Return} = \frac{\text{Closing Price (Day t)}}{\text{Closing Price (Day t-1)}} - 1 $ 2. Compute the 60-day Exponential Moving Average (EMA) of the simple return 3. If the 60-day EMA value is greater than 0, take a long position; otherwise, close the long position [12][13] - Evaluation: The model performed poorly in backtesting, with low win rates (below 30%) and failing to outperform the benchmark indices [13] Trend Return Difference Model - Model Name: Trend Return Difference Model - Construction Idea: The model improves upon the simple return model by introducing the concept of upward and downward return differences to better capture market trends [17][18] - Construction Process: 1. Define upward return as: $ \text{Upward Return} = \frac{\text{Highest Price (Day t) - Opening Price (Day t)}}{\text{Closing Price (Day t)}} $ 2. Define downward return as: $ \text{Downward Return} = \frac{\text{Opening Price (Day t) - Lowest Price (Day t)}}{\text{Closing Price (Day t)}} $ 3. Calculate the upward and downward return difference: $ \text{Upward-Downward Return Difference} = \text{Upward Return} - \text{Downward Return} $ 4. Compute the 60-day EMA of the upward-downward return difference 5. If the 60-day EMA value is greater than 0, take a long position; otherwise, close the long position [17][18] - Evaluation: The model outperformed the simple return model and the benchmark indices in terms of annualized return, Sharpe ratio, and risk control. It is characterized as a mid-term model with an average long position holding period of approximately 3 weeks [18] Turnover Comprehensive Return Difference Model - Model Name: Turnover Comprehensive Return Difference Model - Construction Idea: Combines turnover and upward-downward return difference to enhance trend-following capabilities by assigning higher weights to trends during high turnover periods [26][27] - Construction Process: 1. Define turnover comprehensive return difference as: $ \text{Turnover Comprehensive Return Difference} = \text{Upward-Downward Return Difference} \times \text{Turnover} $ 2. Compute the 60-day EMA of the turnover comprehensive return difference 3. If the 60-day EMA value is greater than 0, take a long position; otherwise, close the long position [27][28] - Evaluation: The model demonstrated superior performance compared to the simple return model and the upward-downward return difference model. It effectively distinguishes market trends and performs better in high turnover scenarios [27][28] Composite Signal Turnover Comprehensive Return Difference Model - Model Name: Composite Signal Turnover Comprehensive Return Difference Model - Construction Idea: Combines the turnover comprehensive return difference signals from both the Hang Seng Index and the Hang Seng China Enterprises Index to eliminate the randomness caused by differences in index composition [32][33] - Construction Process: 1. Define the composite signal: - If either the Hang Seng Index or the Hang Seng China Enterprises Index turnover comprehensive return difference signal indicates a long position, take a long position in the respective index 2. Compute the 60-day EMA of the composite signal 3. If the composite signal's 60-day EMA value is greater than 0, take a long position; otherwise, close the long position [32][33] - Evaluation: The model significantly outperformed the benchmark indices and single-signal turnover comprehensive return difference models, showcasing robust trend-following capabilities [35][36] --- Model Backtesting Results Simple Return Model - Hang Seng Index: Annualized return 1.26%, maximum drawdown 52.96%, Sharpe ratio -0.044 [15][16] - Hang Seng China Enterprises Index: Annualized return 1.91%, maximum drawdown 68.79%, Sharpe ratio 0.034 [15][16] Trend Return Difference Model - Hang Seng Index: Annualized return 4.23%, maximum drawdown 22.98%, Sharpe ratio 0.154 [19][20] - Hang Seng China Enterprises Index: Annualized return 6.15%, maximum drawdown 37.2%, Sharpe ratio 0.267 [19][20] Turnover Comprehensive Return Difference Model - Hang Seng Index: Annualized return 3%, maximum drawdown 28.84%, Sharpe ratio 0.039 [31] - Hang Seng China Enterprises Index: Annualized return 9.73%, maximum drawdown 24.56%, Sharpe ratio 0.47 [31] Composite Signal Turnover Comprehensive Return Difference Model - Hang Seng Index: Annualized return 7.78%, maximum drawdown 23.81%, Sharpe ratio 0.401 [33][36] - Hang Seng China Enterprises Index: Annualized return 10.03%, maximum drawdown 24.63%, Sharpe ratio 0.484 [33][36] Sensitivity Analysis of Composite Signal Turnover Comprehensive Return Difference Model - Hang Seng Index: - 40-day EMA: Annualized return 6.1%, maximum drawdown 26.78%, Sharpe ratio 0.281 [39] - 50-day EMA: Annualized return 7.02%, maximum drawdown 27.44%, Sharpe ratio 0.34 [39] - 60-day EMA: Annualized return 7.78%, maximum drawdown 23.81%, Sharpe ratio 0.401 [39] - 70-day EMA: Annualized return 7.31%, maximum drawdown 27.2%, Sharpe ratio 0.375 [39] - 80-day EMA: Annualized return 6.86%, maximum drawdown 24.9%, Sharpe ratio 0.343 [39] - Hang Seng China Enterprises Index: - 40-day EMA: Annualized return 8.3%, maximum drawdown 26.72%, Sharpe ratio 0.382 [40] - 50-day EMA: Annualized return 8.97%, maximum drawdown 28.88%, Sharpe ratio 0.416 [40] - 60-day EMA: Annualized return 10.03%, maximum drawdown 24.63%, Sharpe ratio 0.484 [40] - 70-day EMA: Annualized return 9.36%, maximum drawdown 29.04%, Sharpe ratio 0.454 [40] - 80-day EMA: Annualized return 9.04%, maximum drawdown 25.04%, Sharpe ratio 0.438 [40]