Quantitative Models and Construction Methods 1. Model Name: Broad-based Index Timing Strategy - **Model Construction Idea**: Utilize macroeconomic variable combinations to predict the future returns of the CSI All Share Index. The strategy involves making long or short decisions based on the predicted values exceeding a threshold[31][34]. - **Model Construction Process**: 1. Combine liquidity and inventory sub-strategies to predict whether the CSI All Share Index will rise. 2. If any predicted value exceeds the threshold (0.6), go long on the CSI All Share Index; otherwise, go short[31]. - **Model Evaluation**: The model effectively captures the impact of macroeconomic variables on the index, providing a systematic approach to timing[34]. 2. Model Name: Dividend Index Timing Strategy - **Model Construction Idea**: Use combinations of inflation and inventory, as well as inventory and credit, to predict the future returns of the Dividend Index. The strategy involves making long or short decisions based on the average predicted values exceeding a threshold[40]. - **Model Construction Process**: 1. Calculate the average predicted value of inflation + inventory and inventory + credit sub-strategies. 2. If the average exceeds the threshold (0.6), go long on the Dividend Index; otherwise, go short[40]. - **Model Evaluation**: The model demonstrates strong defensive characteristics of the Dividend Index, particularly under specific macroeconomic conditions[40]. 3. Model Name: Style Rotation Strategy - **Model Construction Idea**: Leverage macroeconomic factor combinations to predict the future returns of six style indices. Allocate capital to the top two indices with the highest predicted returns[49][54]. - **Model Construction Process**: 1. Use combinations of inflation + inventory and inflation + credit to predict the future returns of six style indices. 2. Smooth the predicted returns and rank them at the end of each month. 3. Allocate capital equally to the top two indices for the next month[54]. - **Model Evaluation**: The strategy effectively captures the differentiated impacts of macroeconomic factors on various styles, providing a robust framework for style rotation[49][54]. --- Model Backtesting Results 1. Broad-based Index Timing Strategy - **Annualized Return**: 15.34% - **Annualized Volatility**: 22.02% - **Sharpe Ratio**: 0.74 - **Maximum Drawdown**: -28.10% - **Excess Return**: 10.31% - **Tracking Error**: 34.16% - **IR**: 0.30 - **Relative Maximum Drawdown**: -50.30%[36]. 2. Dividend Index Timing Strategy - **Annualized Return**: 10.32% - **Annualized Volatility**: 13.74% - **Sharpe Ratio**: 0.75 - **Maximum Drawdown**: -19.92% - **Excess Return**: 7.97% - **Tracking Error**: 9.23% - **IR**: 0.86 - **Relative Maximum Drawdown**: -12.47%[42]. 3. Style Rotation Strategy - **Annualized Return**: 14.79% - **Annualized Volatility**: 21.81% - **Sharpe Ratio**: 0.64 - **Maximum Drawdown**: -45.93% - **Excess Return**: 4.61% - **Tracking Error**: 10.28% - **IR**: 0.52 - **Relative Maximum Drawdown**: -81.71%[59]. --- Quantitative Factors and Construction Methods 1. Factor Name: Macroeconomic Factor Variables - **Factor Construction Idea**: Select significant macroeconomic sub-variables through regression analysis and weight them inversely by their standard deviation over the past year. Use HP filter to adjust for short-term fluctuations and identify long-term trends[2]. - **Factor Construction Process**: 1. Perform regression of macroeconomic indices against broad-based indices and proxy macroeconomic variables. 2. Select sub-variables with significant t-values. 3. Weight the selected variables inversely by their past-year standard deviation. 4. Apply a one-sided HP filter to remove short-term noise and identify long-term trends[2]. - **Factor Evaluation**: The factor construction process effectively integrates macroeconomic trends and states, providing a comprehensive framework for understanding asset price drivers[2]. 2. Factor Name: High-dimensional Macroeconomic Variables - **Factor Construction Idea**: Combine marginal changes and states of macroeconomic variables to address inconsistencies in traditional macroeconomic factor transmission[2][8]. - **Factor Construction Process**: 1. Identify five dimensions of macroeconomic variables: economic prosperity, inflation, interest rates, inventory, and credit. 2. Combine marginal changes and time-series rankings of these variables to construct high-dimensional macroeconomic factors[9]. - **Factor Evaluation**: The high-dimensional approach improves the stability and predictive power of macroeconomic factors, addressing the limitations of single-dimensional indicators[8][9]. --- Factor Backtesting Results 1. Macroeconomic Factor Variables - **Liquidity (Up)**: 70.30% probability of index rise - **Liquidity (Down)**: 58.33% probability of index rise - **Inventory (Up)**: 65.84% probability of index rise - **Inventory (Down)**: 63.91% probability of index rise[37]. 2. High-dimensional Macroeconomic Variables - **Inflation (Up)**: 58.91% probability of index rise - **Inflation (Down)**: 67.33% probability of index rise - **Inventory (Up)**: 64.13% probability of index rise - **Inventory (Down)**: 63.91% probability of index rise[47].

- The report constructs macro factor variables by regressing macro indices on broad-based indices and proxy macro variables, selecting significant sub-macro variables, and weighting them by the inverse of the past year's standard deviation. The macroeconomic data is adjusted using a one-sided HP filter to eliminate short-term fluctuations and identify long-term trends. Based on the filtered variables, macro trends (upward, downward) are divided using factor momentum, and macro states (high, medium, low) are divided using time series percentiles[2] - The necessity of elevating macro factors is highlighted, as the price transmission of macro factor A to broad-based, style, and industry indices varies with different marginal changes of A. Additionally, the impact of macro factor A on returns is different under various states of macro factor B. The combination of marginal changes and states of macro variables is required to comprehensively consider the trend and time series ranking of macro variables[2] - The small-cap all-index timing strategy, based on a combination of macro variables, achieved an annualized return of 16.56% from the end of January 2012 to January 31, 2026, with an excess return of 10.19% relative to the CSI All Index[3] - The dividend index timing strategy, based on a combination of macro variables, achieved an annualized return of 10.97% from the end of January 2012 to January 31, 2026, with an excess return of 8.49% relative to the dividend index[3] - The style rotation strategy, based on a combination of macro variables, achieved an annualized return of 14.79% from September 30, 2014, to January 31, 2026, with an excess return of 4.61% relative to the equal-weighted style index[3] Model and Factor Construction 1. **Macro Factor Variables** - **Construction Idea**: Regress macro indices on broad-based indices and proxy macro variables, select significant sub-macro variables, and weight them by the inverse of the past year's standard deviation[2] - **Construction Process**: - Regress macro indices on broad-based indices and proxy macro variables - Select sub-macro variables with significant t-values - Weight selected variables by the inverse of the past year's standard deviation - Adjust macroeconomic data using a one-sided HP filter to eliminate short-term fluctuations - Divide macro trends using factor momentum and macro states using time series percentiles[2] - **Evaluation**: The necessity of elevating macro factors is highlighted due to the varying price transmission of macro factors under different marginal changes and states[2] 2. **Small-Cap All-Index Timing Strategy** - **Construction Idea**: Use a combination of macro variables to predict the highest returns when inventory is at a medium upward level[3] - **Construction Process**: - Combine macro variables to construct the timing strategy - Evaluate the strategy's performance from January 2012 to January 31, 2026[3] - **Evaluation**: The strategy achieved significant excess returns relative to the CSI All Index[3] 3. **Dividend Index Timing Strategy** - **Construction Idea**: Use a combination of macro variables to construct the dividend index allocation strategy[3] - **Construction Process**: - Combine macro variables to construct the timing strategy - Evaluate the strategy's performance from January 2012 to January 31, 2026[3] - **Evaluation**: The strategy achieved significant excess returns relative to the dividend index[3] 4. **Style Rotation Strategy** - **Construction Idea**: Use a combination of macro variables to construct the style rotation allocation strategy[3] - **Construction Process**: - Combine macro variables to construct the style rotation strategy - Evaluate the strategy's performance from September 30, 2014, to January 31, 2026[3] - **Evaluation**: The strategy achieved significant excess returns relative to the equal-weighted style index[3] Model Backtest Results 1. **Small-Cap All-Index Timing Strategy** - **Annualized Return**: 16.56% - **Excess Return**: 10.19%[3] 2. **Dividend Index Timing Strategy** - **Annualized Return**: 10.97% - **Excess Return**: 8.49%[3] 3. **Style Rotation Strategy** - **Annualized Return**: 14.79% - **Excess Return**: 4.61%[3]

Quantitative Models and Construction - **Model Name**: Volume Model **Construction Idea**: This model uses trading volume as a key indicator to predict short-term market trends[2][12][75] **Construction Process**: The model analyzes the trading volume of broad-based indices to generate "bullish" or "neutral" signals for short-term market timing[12][75] **Evaluation**: The model provides actionable signals for short-term market movements, but its effectiveness may vary depending on market conditions[12][75] - **Model Name**: Institutional Feature Model (Dragon-Tiger List) **Construction Idea**: This model leverages institutional trading data from the Dragon-Tiger List to assess market sentiment[12][75] **Construction Process**: It evaluates institutional trading patterns and generates "neutral" signals for short-term market timing[12][75] **Evaluation**: The model is useful for gauging institutional sentiment but may lack precision in volatile markets[12][75] - **Model Name**: Intelligent Algorithm Model (CSI 300 and CSI 500) **Construction Idea**: This model applies machine learning algorithms to predict market trends for specific indices[12][75] **Construction Process**: The model generates "bullish" signals for the CSI 300 and CSI 500 indices based on algorithmic analysis of historical data[12][75] **Evaluation**: The model demonstrates strong predictive capabilities for these indices, particularly in stable market conditions[12][75] - **Model Name**: Limit-Up/Limit-Down Model **Construction Idea**: This model uses the frequency of limit-up and limit-down events to assess medium-term market trends[13][76] **Construction Process**: It generates "neutral" signals for all broad-based indices by analyzing the distribution of such events over a specific period[13][76] **Evaluation**: The model is effective in identifying market extremes but may not capture subtle trends[13][76] - **Model Name**: Up-Down Return Difference Model **Construction Idea**: This model calculates the difference between upward and downward returns to predict medium-term trends[13][76] **Construction Process**: It generates "bullish" signals for all broad-based indices by analyzing the return asymmetry[13][76] **Evaluation**: The model is robust in identifying directional trends but may lag in rapidly changing markets[13][76] - **Model Name**: Calendar Effect Model **Construction Idea**: This model incorporates seasonal patterns to predict medium-term market movements[13][76] **Construction Process**: It generates "neutral" signals by analyzing historical calendar-based trends[13][76] **Evaluation**: The model is useful for identifying seasonal effects but may not account for external shocks[13][76] - **Model Name**: Long-Term Momentum Model **Construction Idea**: This model uses momentum indicators to predict long-term market trends[14][77] **Construction Process**: It generates "neutral" signals by analyzing long-term price momentum[14][77] **Evaluation**: The model is effective for long-term trend identification but may underperform in choppy markets[14][77] - **Model Name**: A-Share Comprehensive Weapon V3 Model **Construction Idea**: This composite model integrates multiple signals to provide a comprehensive market outlook[15][78] **Construction Process**: It generates "bullish" signals by combining short-term, medium-term, and long-term indicators[15][78] **Evaluation**: The model offers a balanced perspective but may dilute the impact of individual signals[15][78] - **Model Name**: A-Share Comprehensive Guozheng 2000 Model **Construction Idea**: This model focuses on the Guozheng 2000 index using a composite approach[15][78] **Construction Process**: It generates "neutral" signals by integrating various indicators specific to the Guozheng 2000 index[15][78] **Evaluation**: The model is tailored for this index but may lack generalizability[15][78] - **Model Name**: Turnover-to-Volatility Model (Hong Kong Market) **Construction Idea**: This model uses the ratio of turnover to volatility to predict medium-term trends in the Hong Kong market[16][79] **Construction Process**: It generates "bullish" signals by analyzing the turnover-to-volatility ratio[16][79] **Evaluation**: The model is effective in capturing liquidity-driven trends but may not account for external factors[16][79] - **Model Name**: Up-Down Return Similarity Model (Hong Kong Market) **Construction Idea**: This model compares the similarity of upward and downward returns to predict medium-term trends[16][79] **Construction Process**: It generates "bullish" signals for the Hang Seng Index by analyzing return patterns[16][79] **Evaluation**: The model is useful for identifying consistent trends but may struggle in highly volatile markets[16][79] Model Backtesting Results - **Volume Model**: Generates "bullish" signals for specific broad-based indices[12][75] - **Institutional Feature Model**: Generates "neutral" signals for short-term market timing[12][75] - **Intelligent Algorithm Model**: Generates "bullish" signals for CSI 300 and CSI 500 indices[12][75] - **Limit-Up/Limit-Down Model**: Generates "neutral" signals for all broad-based indices[13][76] - **Up-Down Return Difference Model**: Generates "bullish" signals for all broad-based indices[13][76] - **Calendar Effect Model**: Generates "neutral" signals for medium-term trends[13][76] - **Long-Term Momentum Model**: Generates "neutral" signals for long-term trends[14][77] - **A-Share Comprehensive Weapon V3 Model**: Generates "bullish" signals for the overall market[15][78] - **A-Share Comprehensive Guozheng 2000 Model**: Generates "neutral" signals for the Guozheng 2000 index[15][78] - **Turnover-to-Volatility Model**: Generates "bullish" signals for the Hong Kong market[16][79] - **Up-Down Return Similarity Model**: Generates "bullish" signals for the Hang Seng Index[16][79]

Group 1 - The report emphasizes the construction of a high-dimensional macroeconomic factor system to analyze the impact of macroeconomic variables on asset prices and to predict future trends in broad market indices and industry profitability [2][8][9] - It identifies five dimensions of macroeconomic variables: economic prosperity, inflation, interest rates, inventory, and credit, to improve the stability of macroeconomic assessments [9] Group 2 - Current macroeconomic conditions indicate a weak recovery, with overall indicators dropping from 72% to 61%, and industrial output and GDP growth rates remaining flat [17][19] - The report highlights that while inflation remains low, liquidity conditions are stable, and credit indicators show signs of improvement, suggesting a gradual recovery in financing demand [19][20] Group 3 - A broad market timing strategy based on macroeconomic variables has achieved an annualized return of 16.2% from January 2012 to December 2025, outperforming the industry by 10.26% [3][30] - The dividend index timing strategy has yielded an annualized return of 10.78%, exceeding the industry benchmark by 8.42% during the same period [3][37] Group 4 - The style rotation strategy, constructed from macroeconomic variables, has produced an annualized return of 14.15%, outperforming equal-weighted style indices by 6.08% from September 2014 to December 2025 [3][50] - The report suggests maintaining a balanced allocation between dividend and value stocks, while being cautious with growth and performance stocks due to current macroeconomic conditions [23][60]

- The report mentions multiple quantitative models, including the "Volume Model," "Low Volatility Model," "Feature Institutional Model," "Feature Volume Model," "Smart CSI 300 Model," "Smart CSI 500 Model," "Limit-Up/Down Model," "Calendar Effect Model," "Long-Term Momentum Model," "Comprehensive Weaponry V3 Model," and "Comprehensive CSI 2000 Model"[2][12][13][15] - The "Volume Model" is short-term and indicates a bullish signal for most broad-based indices[12] - The "Low Volatility Model" is short-term and provides a neutral signal[12] - The "Feature Institutional Model" is short-term and indicates a bearish signal[12] - The "Feature Volume Model" is short-term and indicates a bullish signal[12] - The "Smart CSI 300 Model" and "Smart CSI 500 Model" are short-term and indicate bullish signals[12] - The "Limit-Up/Down Model" is a mid-term model and provides a neutral signal[13] - The "Calendar Effect Model" is a mid-term model and provides a neutral signal[13] - The "Long-Term Momentum Model" is a long-term model and provides a neutral signal for all broad-based indices[14] - The "Comprehensive Weaponry V3 Model" and "Comprehensive CSI 2000 Model" are composite models that indicate bullish signals[15] - The "Turnover-to-Volatility Model" is a mid-term model for the Hong Kong market and indicates a bullish signal[16] - Backtesting results for the "Double Bottom Pattern" show a weekly return of 2.52%, outperforming the Shanghai Composite Index by 1.43%[43] - Backtesting results for the "Cup-and-Handle Pattern" show a weekly return of 2.0%, outperforming the Shanghai Composite Index by 0.91%[43]

- The short-term A-share models include the volume model (neutral), low volatility model (neutral), characteristic institutional model (neutral), characteristic volume model (bearish), intelligent CSI 300 model (bullish), and intelligent CSI 500 model (bearish) [1][10][68] - The mid-term A-share models include the limit-up and limit-down model (neutral) and the calendar effect model (neutral) [11][69] - The long-term A-share model is the long-term momentum model, which is neutral for all broad-based indices [12][70] - The comprehensive A-share models include the A-share comprehensive weapon V3 model (bearish) and the A-share comprehensive Guozheng 2000 model (bearish) [13][71] - The mid-term Hong Kong stock model is the turnover amplitude model, which is bullish [14][72] - The backtesting results for the models are as follows: - Volume model: neutral [1][10][68] - Low volatility model: neutral [1][10][68] - Characteristic institutional model: neutral [1][10][68] - Characteristic volume model: bearish [1][10][68] - Intelligent CSI 300 model: bullish [1][10][68] - Intelligent CSI 500 model: bearish [1][10][68] - Limit-up and limit-down model: neutral [11][69] - Calendar effect model: neutral [11][69] - Long-term momentum model: neutral [12][70] - A-share comprehensive weapon V3 model: bearish [13][71] - A-share comprehensive Guozheng 2000 model: bearish [13][71] - Turnover amplitude model: bullish [14][72]