Quantitative Models and Construction Methods 1. Model Name: Stock Index Futures Timing Model - **Model Construction Idea**: The model uses the basis of stock index futures to reflect market sentiment changes and constructs daily frequency timing signals based on this correlation[7] - **Model Construction Process**: - The model groups and tests the correlation trend between the basis of stock index futures and the index itself - Constructs daily frequency timing signals based on this correlation - As of October 17, 2025, the timing signal based on the basis of the CSI 500 stock index futures remained at 1[31] - **Model Evaluation**: The model effectively captures market sentiment changes and provides timely signals for trading decisions[7] 2. Model Name: Multi-Dimensional Timing Model - **Model Construction Idea**: The model integrates macro, micro, meso, and derivative signals to form a four-dimensional non-linear timing model[33] - **Model Construction Process**: - The A-share market is divided into nine states based on macro, micro, and meso signals, each corresponding to long and short signals to form a three-dimensional large cycle timing signal - On this basis, the derivative signal generated by the basis of stock index futures is superimposed to form a four-dimensional non-linear timing model - The latest composite multi-dimensional timing signal is long (1)[34] - **Model Evaluation**: The model provides a comprehensive view of market conditions by integrating multiple dimensions, enhancing the accuracy of timing signals[33] 3. Model Name: Style Enhancement Model - **Model Construction Idea**: The model enhances returns by adding enhancement factors to the multi-style strategy, suppressing single-style fluctuations, and achieving stable excess returns in different cycles[41] - **Model Construction Process**: - The model is based on the multi-style strategy and adds enhancement factors - It dynamically adjusts the weights of different styles to achieve stable excess returns - As of October 17, 2025, the low volatility enhancement strategy achieved an excess return of 6.05%[42] - **Model Evaluation**: The model effectively enhances returns while controlling risks, providing stable performance across different market cycles[41] Model Backtesting Results Stock Index Futures Timing Model - **Absolute Return**: Not specified - **Excess Return**: 4.33%[9] - **Annualized Return**: Not specified - **Sharpe Ratio**: Not specified Multi-Dimensional Timing Model - **Absolute Return**: Not specified - **Excess Return**: 4.33%[9] - **Annualized Return**: Not specified - **Sharpe Ratio**: Not specified Style Enhancement Model - **Absolute Return**: Not specified - **Excess Return**: 6.05%[8] - **Annualized Return**: Not specified - **Sharpe Ratio**: Not specified Quantitative Factors and Construction Methods 1. Factor Name: High-Frequency Factor - **Factor Construction Idea**: The factor captures market valuation and sentiment risks using high-frequency data[11] - **Factor Construction Process**: - The factor uses high-frequency data to measure market depth, spread, and price elasticity - Constructs indicators such as average depth, spread, and price elasticity to reflect market liquidity and sentiment - For example, the average depth is calculated as: $$ avg_{depth} = \frac{av1 + bv1}{2} $$ where av1 and bv1 are the sell and buy volumes at the first level of the order book[98] - **Factor Evaluation**: The factor effectively captures market liquidity and sentiment changes, providing valuable insights for trading decisions[11] Factor Backtesting Results High-Frequency Factor - **Absolute Return**: Not specified - **Excess Return**: Not specified - **Annualized Return**: Not specified - **Sharpe Ratio**: Not specified Industry and ETF Rotation Strategy 1. Strategy Name: Industry Rotation Strategy - **Strategy Construction Idea**: The strategy uses quantitative fundamental drivers, quality low volatility style drivers, and distressed reversal industry discovery methods to construct an industry rotation strategy[76] - **Strategy Construction Process**: - Combines industry fundamental rotation, quality low volatility, and distressed reversal three-dimensional industry rotation strategies into an equal-weight portfolio - Selects industries from different dimensions to achieve factor and style complementarity, reducing the risk of a single strategy - As of October 17, 2025, the annualized excess return of the industry rotation strategy based on three-strategy integration was 10.59%, with a Sharpe ratio of 0.74[80] - **Strategy Evaluation**: The strategy effectively combines multiple dimensions to enhance returns while controlling risks, providing stable performance across different market cycles[76] Strategy Backtesting Results Industry Rotation Strategy - **Absolute Return**: Not specified - **Excess Return**: 14.75%[10] - **Annualized Return**: 10.59%[80] - **Sharpe Ratio**: 0.74[80]

二 〇 二 五 年 度 2025 年 04 月 22 日 国 泰 君 安 期 货 研 究 所 李浩 投资咨询从业资格号：Z0020121 lihao8@gtht.com 宋子钰（联系人） 从业资格证号：F03136034 songziyu@gtht.com 报告导读： 在传统的策略中，量价因子（如动量、波动率、成交量等）占据主导地位，而宏观流动性和资金面因 子的在择时中的有效性往往被低估。然而，市场短期走势不仅受技术面影响，更与资金松紧、机构行为和 市场情绪密切相关。1 天期的国债逆回购利率作为反映短期资金供需的核心指标，其高频波动蕴含了丰富 的市场信息——它不仅是流动性的"温度计"，更是机构投资者行为的"显微镜"。 本报告从 1 天期的国债逆回购利率的日内特征出发，探索其在股指期货择时中的应用价值。不同于传 统宏观因子的低频特性，1 天期的国债逆回购利率的分钟级数据能够捕捉市场资金的实时变化，尤其是在 尾盘阶段，其波动往往隐含某些关键性的次日市场预期。通过构建一系列基于 1 天期的国债逆回购利率时 间序列、量价关系及市场微观结构的因子，我们验证了资金面信息对股指期货短期走势的预测能力，并进 一步在股指期货标 ...