证券研究报告 | 金融工程 2026 年 2 月 28 日 利率价量择时观点转向偏空 ——利率市场趋势定量跟踪 20260227 利率市场结构变化 - 10 年期国债到期收益率录得 1.79%，相对上周下降 0.51BP。当前 利率水平、期限和凸性结构读数分别为 1.56%、0.47%、-0.01%， 从均值回归视角看，目前处于水平结构点位较低、期限结构点位 中性偏低、凸性结构点位偏低的状态。 利率价量周期择时信号：5 年期中性、10 年期偏空、30 年期偏空 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： 长周期向下突破、中周期无信号、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 0 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 2.29%、2.64%、2.73%，最大回撤为 0.69%、 0.92%、1.69%，收益回撤比为 3.88、4.57、2.98，相对业绩基准的 超额收益率为 0.63%、1.13%、 ...

Quantitative Models and Construction Methods 1. Model Name: Multi-Cycle Timing Model for Domestic Interest Rates - **Model Construction Idea**: The model uses kernel regression algorithms to identify the support and resistance lines of interest rate trends. It evaluates the breakthrough patterns of interest rate movements across different investment cycles to generate timing signals[10][22][19] - **Model Construction Process**: - **Data Input**: Yield-to-Maturity (YTM) data for 5-year, 10-year, and 30-year government bonds[10][22] - **Cycle Definition**: - Long cycle: Monthly frequency - Medium cycle: Bi-weekly frequency - Short cycle: Weekly frequency[10][22] - **Signal Generation**: - A signal is generated when at least two out of three cycles show consistent directional breakthroughs (upward or downward) - For example, if two cycles show upward breakthroughs, the final signal is "bearish" for interest rates[10][22] - **Model Evaluation**: The model effectively captures multi-cycle resonance in interest rate trends and provides actionable timing signals[10][22] 2. Model Name: Multi-Cycle Timing Model for US Interest Rates - **Model Construction Idea**: The domestic multi-cycle timing model is applied to the US Treasury market to assess its effectiveness in a different market environment[19] - **Model Construction Process**: - **Data Input**: Yield-to-Maturity (YTM) data for 10-year US Treasury bonds[19] - **Cycle Definition**: - Long cycle: Monthly frequency - Medium cycle: Bi-weekly frequency - Short cycle: Weekly frequency[19] - **Signal Generation**: - A signal is generated when at least two out of three cycles show consistent directional breakthroughs - If no consistent breakthroughs are observed, the final signal is "neutral"[19] - **Model Evaluation**: The model demonstrates adaptability to the US market but currently shows no significant signals, indicating a "neutral" stance[19] --- Model Backtesting Results 1. Multi-Cycle Timing Model for Domestic Interest Rates - **5-Year YTM Model**: - Long-term annualized return: 5.46% - Maximum drawdown: 2.88% - Return-to-drawdown ratio: 1.9 - Short-term annualized return (since 2024): 2.04% - Maximum drawdown: 0.59% - Return-to-drawdown ratio: 3.47 - Long-term excess return: 1.06% - Short-term excess return: 0.74%[23][35] - **10-Year YTM Model**: - Long-term annualized return: 6.03% - Maximum drawdown: 2.74% - Return-to-drawdown ratio: 2.2 - Short-term annualized return (since 2024): 2.3% - Maximum drawdown: 0.58% - Return-to-drawdown ratio: 3.98 - Long-term excess return: 1.63% - Short-term excess return: 1.2%[26][35] - **30-Year YTM Model**: - Long-term annualized return: 7.28% - Maximum drawdown: 4.27% - Return-to-drawdown ratio: 1.7 - Short-term annualized return (since 2024): 2.44% - Maximum drawdown: 0.92% - Return-to-drawdown ratio: 2.66 - Long-term excess return: 2.39% - Short-term excess return: 2.29%[31][35] 2. Multi-Cycle Timing Model for US Interest Rates - **10-Year YTM Model**: - Current signal: Neutral - No significant breakthroughs observed in long, medium, or short cycles[19][21] --- Quantitative Factors and Construction Methods 1. Factor Name: Interest Rate Structure Indicators (Level, Term, Convexity) - **Factor Construction Idea**: Transform YTM data into structural indicators to analyze the interest rate market from a mean-reversion perspective[7] - **Factor Construction Process**: - **Level Structure**: - Formula: $ \text{Level} = \text{Average YTM across maturities} $ - Current value: 1.64% - Historical percentiles: 29% (3 years), 17% (5 years), 9% (10 years)[7] - **Term Structure**: - Formula: $ \text{Term} = \text{YTM (10-year)} - \text{YTM (1-year)} $ - Current value: 0.59% - Historical percentiles: 56% (3 years), 41% (5 years), 48% (10 years)[7] - **Convexity Structure**: - Formula: $ \text{Convexity} = \text{Second derivative of YTM curve} $ - Current value: 0.14% - Historical percentiles: 53% (3 years), 32% (5 years), 32% (10 years)[7] - **Factor Evaluation**: These indicators provide a comprehensive view of the interest rate market's structural dynamics and are useful for mean-reversion analysis[7] --- Factor Backtesting Results 1. Interest Rate Structure Indicators - **Level Structure**: Current value: 1.64%, historical percentiles: 29% (3 years), 17% (5 years), 9% (10 years)[7] - **Term Structure**: Current value: 0.59%, historical percentiles: 56% (3 years), 41% (5 years), 48% (10 years)[7] - **Convexity Structure**: Current value: 0.14%, historical percentiles: 53% (3 years), 32% (5 years), 32% (10 years)[7]

Quantitative Models and Construction Methods 1. Model Name: Multi-Cycle Timing Model for Domestic Interest Rates - **Model Construction Idea**: The model uses kernel regression algorithms to identify the support and resistance lines of interest rate trends. It evaluates the breakthrough patterns of interest rate movements across different investment cycles to provide multi-cycle composite timing signals[9][21]. - **Model Construction Process**: 1. **Data Input**: Yield-to-Maturity (YTM) data for 5-year, 10-year, and 30-year government bonds is used as the basis for analysis[9][21]. 2. **Cycle Definition**: - Long cycle: Monthly frequency - Medium cycle: Bi-weekly frequency - Short cycle: Weekly frequency[9][21]. 3. **Signal Generation**: - For each cycle, the model identifies upward or downward breakthroughs in interest rate trends. - A composite score is calculated based on the number of consistent signals across the three cycles. If at least two out of three cycles show the same directional breakthrough, a timing signal is generated[9][21]. 4. **Final Signal**: - The final signal is classified as "neutral," "neutral oscillation," or "neutral with a slight upward/downward bias" based on the composite score[9][21]. - **Model Evaluation**: The model effectively captures interest rate trends and provides actionable timing signals for different bond maturities[9][21]. 2. Model Name: Multi-Cycle Timing Model for U.S. Interest Rates - **Model Construction Idea**: The domestic multi-cycle timing model is applied to the U.S. Treasury market to generate timing signals for U.S. interest rates[18]. - **Model Construction Process**: 1. **Data Input**: 10-year U.S. Treasury YTM data is used as the basis for analysis[18]. 2. **Cycle Definition**: - Long cycle: Monthly frequency - Medium cycle: Bi-weekly frequency - Short cycle: Weekly frequency[18]. 3. **Signal Generation**: - The model identifies upward or downward breakthroughs in interest rate trends for each cycle. - A composite score is calculated based on the number of consistent signals across the three cycles. If at least two out of three cycles show the same directional breakthrough, a timing signal is generated[18]. 4. **Final Signal**: - The final signal is classified as "neutral" or "neutral oscillation" based on the composite score[18]. - **Model Evaluation**: The model demonstrates adaptability to the U.S. market, providing consistent timing signals for U.S. Treasury yields[18]. --- Model Backtesting Results 1. Multi-Cycle Timing Model for Domestic Interest Rates - **5-Year YTM**: - Long-term annualized return: 5.47% - Maximum drawdown: 2.88% - Return-to-drawdown ratio: 1.9 - Short-term annualized return (since 2024): 2.1% - Maximum drawdown: 0.59% - Return-to-drawdown ratio: 3.57 - Long-term excess return: 1.06% - Short-term excess return: 0.77%[22][34] - **10-Year YTM**: - Long-term annualized return: 6.04% - Maximum drawdown: 2.74% - Return-to-drawdown ratio: 2.21 - Short-term annualized return (since 2024): 2.33% - Maximum drawdown: 0.58% - Return-to-drawdown ratio: 4.03 - Long-term excess return: 1.63% - Short-term excess return: 1.19%[25][34] - **30-Year YTM**: - Long-term annualized return: 7.32% - Maximum drawdown: 4.27% - Return-to-drawdown ratio: 1.71 - Short-term annualized return (since 2024): 2.95% - Maximum drawdown: 0.92% - Return-to-drawdown ratio: 3.22 - Long-term excess return: 2.42% - Short-term excess return: 2.65%[30][34] 2. Multi-Cycle Timing Model for U.S. Interest Rates - **10-Year YTM**: - Final signal: Neutral oscillation - Composite score: 1 upward breakthrough in the short cycle, no signals in the long and medium cycles[18][20] --- Quantitative Factors and Construction Methods 1. Factor Name: Interest Rate Structure Indicators (Level, Term, Convexity) - **Factor Construction Idea**: The factor set quantifies the structural characteristics of the interest rate market, including level, term, and convexity, to provide insights into market positioning and mean-reversion tendencies[6]. - **Factor Construction Process**: 1. **Level Structure**: - Definition: Average YTM across maturities - Current value: 1.58% - Historical percentiles: 18% (3 years), 11% (5 years), 5% (10 years)[6]. 2. **Term Structure**: - Definition: Slope of the yield curve - Current value: 0.55% - Historical percentiles: 50% (3 years), 34% (5 years), 40% (10 years)[6]. 3. **Convexity Structure**: - Definition: Curvature of the yield curve - Current value: 0.06% - Historical percentiles: 41% (3 years), 25% (5 years), 21% (10 years)[6]. - **Factor Evaluation**: These factors provide a comprehensive view of the interest rate market's structural dynamics, aiding in timing and allocation decisions[6]. --- Factor Backtesting Results 1. Interest Rate Structure Indicators - **Level Structure**: Current value: 1.58%, historical percentiles: 18% (3 years), 11% (5 years), 5% (10 years)[6] - **Term Structure**: Current value: 0.55%, historical percentiles: 50% (3 years), 34% (5 years), 40% (10 years)[6] - **Convexity Structure**: Current value: 0.06%, historical percentiles: 41% (3 years), 25% (5 years), 21% (10 years)[6]

Quantitative Models and Construction Methods 1. Model Name: Multi-Cycle Timing Model Based on Bond YTM - **Model Construction Idea**: The model uses kernel regression algorithms to capture the trend patterns of interest rates, depicting the support and resistance lines of interest rate data. It provides multi-cycle composite timing views based on the shape breakthrough situations of interest rate trends in different investment cycles[10][18]. - **Model Construction Process**: - **Data Source**: 5-year, 10-year, and 30-year government bond YTM data. - **Signal Calculation**: - Long cycle: monthly frequency - Medium cycle: bi-weekly frequency - Short cycle: weekly frequency - **Signal Interpretation**: - 5-year bond YTM: Long cycle no signal, medium cycle downward breakthrough, short cycle no signal. Final signal: neutral oscillation[10][12]. - 10-year bond YTM: Long cycle downward breakthrough, medium cycle downward breakthrough, short cycle downward breakthrough. Final signal: bullish[13][15]. - 30-year bond YTM: Long cycle no signal, medium cycle downward breakthrough, short cycle downward breakthrough. Final signal: bullish[16][17]. - **Model Evaluation**: The model effectively captures the trend patterns of interest rates and provides clear timing signals based on multi-cycle analysis. Model Backtesting Results 1. Multi-Cycle Timing Model Based on 5-Year Bond YTM - **Annualized Return**: 2.15% since the end of 2024[4][22]. - **Maximum Drawdown**: 0.73%[4]. - **Return-Drawdown Ratio**: 3.64[4]. - **Excess Return**: 0.8% relative to the performance benchmark[4]. - **Probability of Positive Annual Absolute Return**: Close to 100% since 2008[4]. 2. Multi-Cycle Timing Model Based on 10-Year Bond YTM - **Annualized Return**: 2.47% since the end of 2024[4][25]. - **Maximum Drawdown**: 0.98%[4]. - **Return-Drawdown Ratio**: 4.28[4]. - **Excess Return**: 1.32% relative to the performance benchmark[4]. - **Probability of Positive Annual Absolute Return**: Close to 100% since 2008[4]. 3. Multi-Cycle Timing Model Based on 30-Year Bond YTM - **Annualized Return**: 3.07% since the end of 2024[4][30]. - **Maximum Drawdown**: 1.72%[4]. - **Return-Drawdown Ratio**: 3.35[4]. - **Excess Return**: 2.78% relative to the performance benchmark[4]. - **Probability of Positive Annual Absolute Return**: Close to 100% since 2008[4].

Quantitative Models and Construction Methods - **Model Name**: Multi-cycle timing model for domestic interest rate price-volume trends **Model Construction Idea**: The model uses kernel regression algorithms to capture interest rate trend patterns, identifying support and resistance lines based on the shape of interest rate movements across different investment cycles [10][24] **Model Construction Process**: 1. **Data Input**: Utilize 5-year, 10-year, and 30-year government bond YTM data as the basis for analysis [10][24] 2. **Cycle Classification**: Divide the investment horizon into long-term (monthly frequency), medium-term (bi-weekly frequency), and short-term (weekly frequency) cycles [10][24] 3. **Signal Identification**: Detect upward or downward breakthroughs of support and resistance lines for each cycle [10][24] 4. **Composite Scoring**: Aggregate signals across cycles, assigning scores based on the number of consistent breakthroughs (e.g., 2/3 consistent signals lead to a "buy" or "sell" recommendation) [10][24] **Model Evaluation**: The model effectively captures multi-cycle resonance in interest rate trends, providing actionable timing signals for bond trading strategies [10][24] - **Model Name**: Multi-cycle timing model for U.S. interest rate price-volume trends **Model Construction Idea**: Apply the domestic interest rate price-volume timing model to the U.S. Treasury market [21] **Model Construction Process**: 1. **Data Input**: Use 10-year U.S. Treasury YTM data for analysis [21] 2. **Cycle Classification**: Similar to the domestic model, divide the investment horizon into long-term, medium-term, and short-term cycles [21] 3. **Signal Identification**: Detect upward or downward breakthroughs of support and resistance lines for each cycle [21] 4. **Composite Scoring**: Aggregate signals across cycles, assigning scores based on the number of consistent breakthroughs [21] **Model Evaluation**: The model provides a neutral-to-bullish outlook for U.S. Treasury yields, indicating its adaptability to international markets [21] Model Backtesting Results - **Domestic Multi-cycle Timing Model**: - **5-year YTM**: - Long-term annualized return: 5.5% - Maximum drawdown: 2.88% - Return-to-drawdown ratio: 1.91 - Short-term annualized return (since 2024): 1.86% - Maximum drawdown: 0.59% - Return-to-drawdown ratio: 3.16 - Long-term excess return: 1.07% - Short-term excess return: 0.85% [25][27] - **10-year YTM**: - Long-term annualized return: 6.09% - Maximum drawdown: 2.74% - Return-to-drawdown ratio: 2.22 - Short-term annualized return (since 2024): 2.42% - Maximum drawdown: 0.58% - Return-to-drawdown ratio: 4.19 - Long-term excess return: 1.66% - Short-term excess return: 1.55% [28][32] - **30-year YTM**: - Long-term annualized return: 7.38% - Maximum drawdown: 4.27% - Return-to-drawdown ratio: 1.73 - Short-term annualized return (since 2024): 3.11% - Maximum drawdown: 0.92% - Return-to-drawdown ratio: 3.39 - Long-term excess return: 2.42% - Short-term excess return: 2.87% [33][35] - **U.S. Multi-cycle Timing Model**: - **10-year YTM**: - Current signal: Neutral-to-bullish - Long-term annualized return: Not provided - Maximum drawdown: Not provided - Return-to-drawdown ratio: Not provided [21][23] Quantitative Factors and Construction Methods - **Factor Name**: Interest rate structure indicators (level, term, convexity) **Factor Construction Idea**: Transform YTM data into structural indicators to analyze the interest rate market from a mean-reversion perspective [7] **Factor Construction Process**: 1. **Level Structure**: Calculate the average YTM across maturities (1-10 years) 2. **Term Structure**: Measure the slope between short-term and long-term YTM 3. **Convexity Structure**: Assess the curvature of the yield curve [7] **Factor Evaluation**: The indicators effectively capture the current state of the interest rate market, highlighting deviations from historical averages [7] Factor Backtesting Results - **Interest Rate Structure Indicators**: - **Level Structure**: Current reading: 1.64%, historical 10-year percentile: 7% - **Term Structure**: Current reading: 0.38%, historical 10-year percentile: 16% - **Convexity Structure**: Current reading: -0.09%, historical 10-year percentile: 1% [7]

Quantitative Models and Construction Methods - **Model Name**: Multi-cycle timing model for domestic interest rate price-volume trends **Model Construction Idea**: The model uses kernel regression algorithms to capture interest rate trend patterns, identifying support and resistance lines based on different investment cycles. It provides composite timing signals by analyzing the shape of interest rate movements across long, medium, and short cycles[10][24][29] **Model Construction Process**: 1. **Data Input**: Use 5-year, 10-year, and 30-year government bond YTM data as the basis for analysis[10][24][29] 2. **Cycle Definition**: Define long, medium, and short cycles with average switching frequencies of monthly, bi-weekly, and weekly, respectively[10][24][29] 3. **Signal Generation**: - If at least two cycles show downward breakthroughs of the support line and the interest rate trend is not upward, allocate fully to long-duration bonds - If at least two cycles show downward breakthroughs of the support line but the interest rate trend is upward, allocate 50% to medium-duration bonds and 50% to long-duration bonds - If at least two cycles show upward breakthroughs of the resistance line and the interest rate trend is not downward, allocate fully to short-duration bonds - If at least two cycles show upward breakthroughs of the resistance line but the interest rate trend is downward, allocate 50% to medium-duration bonds and 50% to short-duration bonds - In other cases, allocate equally across short, medium, and long durations[24][29][29] **Model Evaluation**: The model demonstrates strong adaptability across different market environments and provides consistent timing signals based on multi-cycle resonance[10][24][29] - **Model Name**: Multi-cycle timing model for U.S. interest rate price-volume trends **Model Construction Idea**: The domestic price-volume timing model is applied to the U.S. interest rate market, analyzing long, medium, and short cycles to generate composite timing signals[21][23][24] **Model Construction Process**: 1. **Data Input**: Use 10-year U.S. Treasury YTM data for analysis[21][23][24] 2. **Cycle Definition**: Define long, medium, and short cycles with average switching frequencies of monthly, bi-weekly, and weekly, respectively[21][23][24] 3. **Signal Generation**: Similar to the domestic model, signals are generated based on the number of cycles showing breakthroughs of support or resistance lines and the direction of interest rate trends[21][23][24] **Model Evaluation**: The model effectively captures U.S. interest rate trends and provides reliable timing signals for investment decisions[21][23][24] Model Backtesting Results - **Domestic Multi-cycle Timing Model** - **5-year YTM**: - Long-term annualized return: 5.5% - Maximum drawdown: 2.88% - Return-to-drawdown ratio: 1.91 - Short-term annualized return (since 2024): 1.86% - Maximum drawdown: 0.59% - Return-to-drawdown ratio: 3.15 - Long-term excess return: 1.07% - Excess return-to-drawdown ratio: 0.62 - Short-term excess return: 0.86% - Excess return-to-drawdown ratio: 2.18[25][27][37] - **10-year YTM**: - Long-term annualized return: 6.09% - Maximum drawdown: 2.74% - Return-to-drawdown ratio: 2.23 - Short-term annualized return (since 2024): 2.35% - Maximum drawdown: 0.58% - Return-to-drawdown ratio: 4.07 - Long-term excess return: 1.66% - Excess return-to-drawdown ratio: 1.16 - Short-term excess return: 1.56% - Excess return-to-drawdown ratio: 3.46[28][32][37] - **30-year YTM**: - Long-term annualized return: 7.38% - Maximum drawdown: 4.27% - Return-to-drawdown ratio: 1.73 - Short-term annualized return (since 2024): 2.98% - Maximum drawdown: 0.92% - Return-to-drawdown ratio: 3.26 - Long-term excess return: 2.42% - Excess return-to-drawdown ratio: 0.87 - Short-term excess return: 2.87% - Excess return-to-drawdown ratio: 3.21[33][35][37] - **U.S. Multi-cycle Timing Model** - **10-year YTM**: - Composite signal: Long cycle upward breakthrough, medium and short cycles downward breakthrough - Final signal: Bullish[21][23][24]