国泰海通证券研究·2025-08-01 09:44Core Viewpoint - The article emphasizes the importance of optimizing position strategies in quantitative frameworks for predicting bond futures, rather than solely focusing on the prediction accuracy of price movements [1][3]. Group 1: Position Strategy Optimization - The study tests various position strategies, including a full position strategy as a benchmark, a threshold-based full position strategy, and a gradual accumulation strategy that incorporates a fuzzy interval filtering mechanism [1][3]. - Continuous trading strategies convert binary probability signals into position adjustment signals, allowing for categorization based on risk preferences, such as risk-seeking, risk-averse, and risk-neutral types [1][3]. Group 2: Model and Market Conditions - The report references a multi-factor model for bond market timing, utilizing recent data to train models for predicting the next trading day, with specific market conditions defined for 2024 and 2025 [2]. - The combination of various position strategies is crucial, particularly in volatile markets, where appropriate strategy selection can significantly enhance overall model performance [3]. Group 3: Performance Insights - Binary full position strategies effectively capture returns during clear trends but come with higher volatility and transaction costs [3]. - Gradual accumulation strategies show lower trading frequency advantages, reducing transaction costs, but may have limited return capture in sideways markets [3]. - Single continuous strategies demonstrate strong performance in volatile markets, with specific strategies like Sigmoid and Atanh showing significant advantages in volatility control, especially for risk-averse investors [3].