证券研究报告 | 金融工程 2025 年 11 月 23 日 利率价量择时观点整体转为偏空 美债价量周期择时信号：看多 - 基于美国市场 10 年期国债 YTM 数据判断的多周期择时信号为： 长周期向上突破、中周期向下突破、短周期向下突破。综合来看， 当前合计下行突破 2 票、上行突破 1 票，最终信号的综合评分结 果为看多。 国内利率价量多周期择时策略表现 - 自 2024 年底以来，基于 5/10/30 年期国债 YTM 价量趋势的交易策 略年化收益率分别为 2.24%、2.69%、3.25%，最大回撤为 0.75%、 1.01%、1.78%，收益回撤比为 3.8、4.65、3.54，相对业绩基准的 超额收益率为 0.81%、1.39%、2.57%。2008 年以来，各策略逐年 绝对收益和超额收益大于 0 的概率均接近 100%。 风险提示：本报告基于对历史数据的分析，当市场环境变化时，存在失效 风险。 专题报告 王武蕾 S1090519080001 wangwulei@cmschina.com.cn 梁雨辰 S1090523070008 liangyuchen2@cmschina.com.cn 敬请阅读 ...

- The report introduces a multi-cycle timing strategy for interest rates, which is constructed using shape recognition algorithms to identify support and resistance lines in interest rate trends. The strategy combines signals from short, medium, and long cycles to form composite timing views. The switching frequency for these cycles is weekly, bi-weekly, and monthly, respectively[10][23][24] - The multi-cycle timing strategy is based on the principle that when at least two cycles show downward breakthroughs of support lines and the interest rate trend is not upward, the portfolio is fully allocated to long-duration bonds. Conversely, when at least two cycles show upward breakthroughs of resistance lines and the interest rate trend is not downward, the portfolio is fully allocated to short-duration bonds. Other configurations include mixed allocations depending on the direction of the interest rate trend[23] - The strategy employs a stop-loss mechanism where the portfolio is adjusted to equal-weighted allocation if the daily excess return falls below -0.5%[23] - The backtesting results of the multi-cycle timing strategy show a long-term annualized return of 6.17% since 2007, with a maximum drawdown of 1.52% and a return-to-drawdown ratio of 2.26. Short-term results since the end of 2023 indicate an annualized return of 7.24%, a maximum drawdown of 1.55%, and a return-to-drawdown ratio of 6.21[23][24] - The strategy has consistently outperformed its benchmark, which is an equal-weighted duration strategy, with a long-term excess return of 1.65% and a short-term excess return of 2.14% since the end of 2023. The excess return-to-drawdown ratio is 1.17 for the long term and 2.29 for the short term[23][24] - Historical performance analysis reveals that the strategy achieved a 100% success rate in generating positive absolute returns and excess returns annually over the past 18 years[24] - The report also tracks the behavior of public bond funds using an improved regression model to estimate the duration and divergence of medium- to long-term pure bond funds. The latest results show that the median duration of public bond funds, including leverage, is 3.51 years, with a 4-week moving average of 3.45 years. This represents an increase of 0.13 years and 0.04 years compared to the previous week, respectively, and places the duration level at the 96.53% percentile over the past five years[6][13][14] - The divergence in public bond fund duration, measured by the cross-sectional standard deviation, is 1.55 years, which is slightly lower than the previous week and is at the 59.07% percentile over the past five years[6][14] - The yield-to-maturity (YTM) data for public bond funds, calculated similarly, shows a median YTM of 1.7%, a 4-week moving average of 1.74%, and an average of 1.79%. Compared to the previous week, the unsmoothed median YTM decreased by 4 basis points, while the smoothed data decreased by 3 basis points, indicating that institutional holdings are near historical lows[18]

Quantitative Models and Construction Methods 1. Model Name: Interest Rate Price-Volume Multi-Cycle Timing Strategy - **Model Construction Idea**: This model uses kernel regression algorithms to identify support and resistance levels in interest rate trends. It integrates signals from long, medium, and short investment cycles to form a composite timing strategy[10][23]. - **Model Construction Process**: 1. **Signal Identification**: - Use kernel regression to capture the shape of interest rate trends and identify support and resistance levels[10]. - Classify signals into long-cycle (monthly frequency), medium-cycle (bi-weekly frequency), and short-cycle (weekly frequency)[10]. 2. **Signal Aggregation**: - Count the number of upward and downward breakthroughs across the three cycles. - If at least two cycles show the same directional breakthrough, the composite signal is determined based on the majority[10]. 3. **Portfolio Construction**: - Allocate assets based on the composite signal: - Full allocation to long-duration bonds if at least two cycles show downward breakthroughs and the trend is not upward. - Equal allocation to medium- and long-duration bonds if at least two cycles show downward breakthroughs but the trend is upward. - Full allocation to short-duration bonds if at least two cycles show upward breakthroughs and the trend is not downward. - Equal allocation to medium- and short-duration bonds if at least two cycles show upward breakthroughs but the trend is downward. - Equal allocation across short-, medium-, and long-duration bonds in other cases[23]. 4. **Stop-Loss Mechanism**: - Adjust holdings to equal allocation if the daily excess return of the portfolio falls below -0.5%[23]. 5. **Benchmark**: - The benchmark is an equal-duration strategy with one-third allocation to short-, medium-, and long-duration bonds[23]. - **Model Evaluation**: The model effectively captures multi-cycle resonance in interest rate trends and provides a systematic approach to timing strategies[23]. --- Model Backtesting Results 1. Interest Rate Price-Volume Multi-Cycle Timing Strategy - **Long-Term Performance (2007.12.31 to Latest Report Date)**: - Annualized Return: 6.19% - Maximum Drawdown: 1.53% - Return-to-Drawdown Ratio: 2.26 - Excess Annualized Return: 1.67% - Excess Return-to-Drawdown Ratio: 1.18[23][24] - **Short-Term Performance (Since 2023 Year-End)**: - Annualized Return: 7.5% - Maximum Drawdown: 1.61% - Return-to-Drawdown Ratio: 6.43 - Excess Annualized Return: 2.35% - Excess Return-to-Drawdown Ratio: 2.47[23][24] - **Historical Success Rates (18 Years)**: - Absolute Return > 0: 100% - Excess Return > 0: 100%[24] - **Year-by-Year Performance**: - 2008: Absolute Return 17.08%, Excess Return 4.41% - 2009: Absolute Return 1.03%, Excess Return 1.20% - 2010: Absolute Return 4.59%, Excess Return 2.49% - 2011: Absolute Return 7.25%, Excess Return 2.10% - 2012: Absolute Return 4.33%, Excess Return 0.68% - 2013: Absolute Return 0.91%, Excess Return 1.67% - 2014: Absolute Return 13.47%, Excess Return 2.67% - 2015: Absolute Return 11.14%, Excess Return 2.31% - 2016: Absolute Return 3.20%, Excess Return 1.76% - 2017: Absolute Return 1.11%, Excess Return 1.38% - 2018: Absolute Return 11.16%, Excess Return 2.36% - 2019: Absolute Return 6.24%, Excess Return 1.44% - 2020: Absolute Return 3.46%, Excess Return 0.47% - 2021: Absolute Return 5.40%, Excess Return 0.33% - 2022: Absolute Return 3.62%, Excess Return 0.47% - 2023: Absolute Return 4.81%, Excess Return 0.46% - 2024: Absolute Return 9.35%, Excess Return 2.52% - 2025: Absolute Return 1.14%, Excess Return 0.75%[24][27] --- Quantitative Factors and Construction Methods 1. Factor Name: Interest Rate Structural Indicators (Level, Term, Convexity) - **Factor Construction Idea**: These factors decompose the yield-to-maturity (YTM) data of government bonds into three structural dimensions: level, term, and convexity. The factors are analyzed from a mean-reversion perspective[7][9]. - **Factor Construction Process**: 1. **Data Transformation**: - Convert the YTM data of 1- to 10-year government bonds into three structural indicators: - **Level**: Average YTM across all maturities - **Term**: Difference between long-term and short-term YTM - **Convexity**: Curvature of the yield curve[7]. 2. **Historical Percentile Analysis**: - Calculate the rolling 3-, 5-, and 10-year percentiles for each structural indicator to assess their relative positions[8][9]. - **Factor Evaluation**: These factors provide insights into the current state of the interest rate market and its deviation from historical norms[7][9]. --- Factor Backtesting Results 1. Interest Rate Structural Indicators - **Level**: - Current Value: 1.58% - Weekly Change: -0.24BP - Historical Percentiles: 10% (3-Year), 6% (5-Year), 3% (10-Year)[9] - **Term**: - Current Value: 0.27% - Weekly Change: +4.42BP - Historical Percentiles: 7% (3-Year), 4% (5-Year), 8% (10-Year)[9] - **Convexity**: - Current Value: -0.04% - Weekly Change: -6.28BP - Historical Percentiles: 8% (3-Year), 5% (5-Year), 5% (10-Year)[9]