证券研究报告 | 金融工程 gszqdatemark 2025 12 05 年 月 日 量化点评报告 行业 ETF 配置模型 2025 年超额 16.4% 根据景气度-趋势-拥挤度框架，我们有"强趋势－低拥挤"和"高景气－ 强趋势"两套行业轮动方案。趋势拥挤模型本月推荐农业、传媒、建材、 轻工行业；景气趋势模型本月推荐非银、计算机、家电、煤炭等行业。相 较于上个月，模型明显由高切低，增配金融周期板块。 ① 行业主线模型：相对强弱指标。2024 年出现 RS>90 的行业有：煤炭、 电力及公用事业、家电、银行、石油石化、通信、有色金属、农林牧渔和 汽车。经过全年验证，这些行业确实阶段性地成为了市场的行情主线，上 半年主要是高股息、资源品和出海，下半年主要是 AI。截止到 2025 年 4 月底，共 17 个行业出现 RS>90 的信号，以 TMT 板块、银行、制造和部 分消费行业为主。今年 4 月底前几乎半数行业曾表现强势，行业主线判断 难度较大，四季度配置建议以均衡为主，目前与行业轮动模型选出行业基 本重合。 ② 行业轮动模型：景气度-趋势-拥挤度框架。12 月行业配置组合权重为： 传媒 16%、农林牧渔 1 ...

Quantitative Models and Construction Methods 1. Model Name: Industry Mainline Model (Relative Strength Index, RSI) - **Model Construction Idea**: This model identifies leading industries by calculating their relative strength (RS) based on historical price performance. Industries with RS > 90% are considered potential leaders for the year [10] - **Model Construction Process**: 1. Use 29 first-level industry indices as the investment universe [10] 2. Calculate the price change over the past 20, 40, and 60 trading days for each industry index [10] 3. Rank the price changes for each period and normalize the rankings to obtain RS_20, RS_40, and RS_60 [10] 4. Compute the average of the three rankings to derive the final relative strength index: $ RS = (RS_{20} + RS_{40} + RS_{60}) / 3 $ where RS_20, RS_40, and RS_60 represent the normalized rankings of price changes over 20, 40, and 60 trading days, respectively [10] - **Model Evaluation**: The model successfully identified leading industries in 2024, such as coal, banking, and AI-related sectors, which showed strong performance during the year [10][12] 2. Model Name: Industry Rotation Model (Prosperity-Trend-Crowding Framework) - **Model Construction Idea**: This model combines three dimensions—prosperity, trend, and crowding—to recommend industry allocations. It includes two sub-strategies: "Strong Trend-Low Crowding" and "High Prosperity-Strong Trend" [7][15] - **Model Construction Process**: 1. Define prosperity as the core metric, supplemented by trend and crowding dimensions [15] 2. For the "High Prosperity-Strong Trend" strategy, focus on industries with high prosperity and strong trends while avoiding highly crowded industries [15] 3. For the "Strong Trend-Low Crowding" strategy, prioritize industries with strong trends and low crowding while avoiding low-prosperity industries [15] 4. Allocate weights to industries based on the framework, e.g., November 2025 allocation: Basic Chemicals (18%), Media (16%), Agriculture (12%), Light Manufacturing (12%), Computers (12%), Home Appliances (9%), Real Estate (9%), Retail (6%), New Energy (4%), Coal (3%) [7][15] - **Model Evaluation**: The model demonstrated strong performance, with an annualized excess return of 13.7% and an IR of 1.5. It also showed a high monthly win rate of 67% [15][22] 3. Model Name: Left-Side Inventory Reversal Model - **Model Construction Idea**: This model identifies industries in a recovery phase from distress by analyzing inventory levels and analyst expectations. It aims to capture reversal opportunities in industries with low inventory pressure and potential for restocking [29] - **Model Construction Process**: 1. Focus on industries experiencing current or past distress with signs of recovery [29] 2. Identify industries with low inventory pressure and restocking potential [29] 3. Incorporate analyst long-term positive outlooks for these industries [29] - **Model Evaluation**: The model achieved an absolute return of 27.9% and an excess return of 7.5% relative to equal-weighted industry benchmarks in 2025 (up to October) [29] --- Model Backtesting Results 1. Industry Mainline Model (RSI) - Annualized excess return: Not explicitly stated - IR: Not explicitly stated - Maximum drawdown: Not explicitly stated - Monthly win rate: Not explicitly stated - 2024 performance: Identified leading industries such as coal, banking, and AI, which showed strong performance during the year [10][12] 2. Industry Rotation Model (Prosperity-Trend-Crowding Framework) - Annualized excess return: 13.7% [15] - IR: 1.5 [15] - Maximum drawdown: -8.0% [15] - Monthly win rate: 67% [15] - 2023 excess return: 7.3% [15] - 2024 excess return: 5.7% [15] - 2025 excess return (up to October): 2.0% [15] 3. Left-Side Inventory Reversal Model - Annualized excess return: Not explicitly stated - IR: Not explicitly stated - Maximum drawdown: Not explicitly stated - Monthly win rate: Not explicitly stated - 2023 performance: Absolute return of 13.4%, excess return of 17.0% [29] - 2024 performance: Absolute return of 26.5%, excess return of 15.4% [29] - 2025 performance (up to October): Absolute return of 27.9%, excess return of 7.5% [29]

Quantitative Models and Construction Methods 1. Model Name: Industry Relative Strength (RSI) Model - **Model Construction Idea**: This model identifies leading industries by calculating their relative strength (RS) based on historical price performance over different time windows [10] - **Model Construction Process**: 1. Use 29 first-level industry indices as the configuration targets [10] 2. Calculate the price change rates for the past 20, 40, and 60 trading days for each industry index [10] 3. Rank the industries based on their price change rates for each time window and normalize the rankings to obtain RS_20, RS_40, and RS_60 [10] 4. Calculate the average of the three rankings to derive the final RS value: $ RS = \frac{RS_{20} + RS_{40} + RS_{60}}{3} $ [10] 5. Industries with RS > 90% by the end of April are identified as potential leading industries for the year [10] - **Model Evaluation**: The model successfully identified key annual industry trends, such as high dividend, resource products, exports, and AI, which were validated by market performance throughout the year [10][12] 2. Model Name: Industry Sentiment-Trend-Crowding Framework - **Model Construction Idea**: This framework provides two industry rotation strategies based on market conditions: 1. High sentiment + strong trend, avoiding high crowding (aggressive strategy) 2. Strong trend + low crowding, avoiding low sentiment (conservative strategy) [6][14] - **Model Construction Process**: 1. Evaluate industries based on three dimensions: sentiment, trend, and crowding [6][14] 2. Use sentiment as the core metric for the aggressive strategy, with crowding as a risk control factor [14] 3. Use trend as the core metric for the conservative strategy, avoiding low-sentiment industries [14] 4. Allocate weights to industries based on their scores in the three dimensions [6][14] - **Model Evaluation**: The framework is effective in adapting to different market conditions and has shown strong performance in historical backtests [6][14] 3. Model Name: Left-Side Inventory Reversal Model - **Model Construction Idea**: This model identifies industries with potential for recovery by analyzing sectors in distress or those with low inventory pressure and high analyst optimism [24] - **Model Construction Process**: 1. Identify industries currently in distress or recovering from past distress [24] 2. Focus on sectors with low inventory pressure and potential for restocking [24] 3. Incorporate analyst long-term positive outlooks for these industries [24] - **Model Evaluation**: The model effectively captures recovery opportunities in industries undergoing inventory restocking cycles, providing significant absolute and relative returns [24] --- Model Backtesting Results 1. Industry Relative Strength (RSI) Model - **Annualized Return**: Not explicitly mentioned - **Excess Return**: Not explicitly mentioned - **Information Ratio (IR)**: Not explicitly mentioned - **Maximum Drawdown**: Not explicitly mentioned - **Monthly Win Rate**: Not explicitly mentioned - **Performance Highlights**: - Industries with RS > 90% by April 2024 included coal, utilities, home appliances, banking, petrochemicals, communication, non-ferrous metals, agriculture, and automotive [10] - These industries showed strong performance, with key themes being high dividends, resource products, exports, and AI [10][12] 2. Industry Sentiment-Trend-Crowding Framework - **Annualized Return**: 22.1% (long-only portfolio) [14] - **Excess Return**: 13.8% (annualized) [14] - **Information Ratio (IR)**: 1.51 [14] - **Maximum Drawdown**: -8.0% [14] - **Monthly Win Rate**: 68% [14] - **Performance Highlights**: - 2023 excess return: 7.3% [14] - 2024 excess return: 5.7% [14] - 2025 YTD excess return: 2.8% [14] 3. Left-Side Inventory Reversal Model - **Annualized Return**: Not explicitly mentioned - **Excess Return**: - 2023: 17.0% (relative to equal-weighted industry benchmark) [24] - 2024: 15.4% (relative to equal-weighted industry benchmark) [24] - 2025 YTD: 7.8% (relative to equal-weighted industry benchmark) [24] - **Information Ratio (IR)**: Not explicitly mentioned - **Maximum Drawdown**: Not explicitly mentioned - **Monthly Win Rate**: Not explicitly mentioned - **Performance Highlights**: - Absolute return: - 2023: 13.4% [24] - 2024: 26.5% [24] - 2025 YTD: 26.4% [24] --- Quantitative Factors and Construction Methods 1. Factor Name: Sentiment Factor - **Factor Construction Idea**: Measures the overall sentiment of an industry to identify high-growth opportunities [14] - **Factor Construction Process**: 1. Evaluate the sentiment of each industry based on relevant metrics (not explicitly detailed in the report) [14] 2. Rank industries by sentiment scores [14] - **Factor Evaluation**: Sentiment is a core metric in the aggressive strategy of the Industry Sentiment-Trend-Crowding Framework, providing strong signals for high-growth opportunities [14] 2. Factor Name: Trend Factor - **Factor Construction Idea**: Measures the strength of market trends to identify industries with strong momentum [14] - **Factor Construction Process**: 1. Evaluate the trend of each industry based on relevant metrics (not explicitly detailed in the report) [14] 2. Rank industries by trend scores [14] - **Factor Evaluation**: Trend is a core metric in the conservative strategy of the Industry Sentiment-Trend-Crowding Framework, offering a simple and replicable approach to industry allocation [14] 3. Factor Name: Crowding Factor - **Factor Construction Idea**: Measures the level of crowding in an industry to identify overbought or underbought sectors [14] - **Factor Construction Process**: 1. Evaluate the crowding level of each industry based on relevant metrics (not explicitly detailed in the report) [14] 2. Rank industries by crowding scores [14] - **Factor Evaluation**: Crowding is used as a risk control factor in both aggressive and conservative strategies of the Industry Sentiment-Trend-Crowding Framework [14] --- Factor Backtesting Results 1. Sentiment Factor - **Annualized Return**: Not explicitly mentioned - **Excess Return**: Not explicitly mentioned - **Information Ratio (IR)**: Not explicitly mentioned - **Maximum Drawdown**: Not explicitly mentioned - **Monthly Win Rate**: Not explicitly mentioned 2. Trend Factor - **Annualized Return**: Not explicitly mentioned - **Excess Return**: Not explicitly mentioned - **Information Ratio (IR)**: Not explicitly mentioned - **Maximum Drawdown**: Not explicitly mentioned - **Monthly Win Rate**: Not explicitly mentioned 3. Crowding Factor - **Annualized Return**: Not explicitly mentioned - **Excess Return**: Not explicitly mentioned - **Information Ratio (IR)**: Not explicitly mentioned - **Maximum Drawdown**: Not explicitly mentioned - **Monthly Win Rate**: Not explicitly mentioned