证券研究报告 建议择机入场 2025 年 11 月 23 日│中国内地 量化投资周报 本周观点：建议择机入场 上周，受全球流动性压力、美联储降息预期反复以及 AI 叙事松动多重因素 影响，全球风险偏好下降——VIX 指数攀升至近三个月高位，各类风险资产 均承压，其中比特币、微盘股等对流动性和风偏更敏感的资产领跌。我们的 模型认为 A 股经过上周的调整，整体上消化了过高的估值，观点由防御转 为看平。叠加周五美联储释放了略积极的降息信号，Nowcasting 模型预测 11 月 CPI 或将继续上行至 3.7%-3.8%，但核心 CPI 预计保持平稳，或有利 于市场风偏的恢复。建议择机入场，优选低位防御板块，本周行业轮动模型 加大了对低位消费板块的押注，风格上仍看好红利。 A 股大盘择时模型：上周回调消化了高估值压力，可择机入场 我们以万得全 A 指数作为 A 股大盘代理，从估值、情绪、资金、技术四个 维度对 A 股大盘进行整体方向性判断。今年以来，模型多空择时的扣费后 收益 43.84%，同期 A 股大盘涨跌幅为 20.09%，超额收益为 23.76%；上周 模型超额收益为 10.41%。上周，受全球流动性压力 ...

Group 1 - Market sentiment score has continued to rise, reaching 3 as of November 7, up from 2.7 the previous week, indicating further recovery in market sentiment and a bullish outlook [7][11][19] - The trading volatility between industries has increased rapidly, breaking through the upper Bollinger Band, suggesting accelerated sector switching and a short-term improvement in sentiment [19][22] - The average daily trading volume for the entire A-share market decreased slightly to 20,123.50 billion yuan, with the highest trading day on November 3 at 21,329.04 billion yuan [14][18] Group 2 - The short-term trend scores for industries such as banking, petrochemicals, light manufacturing, electric equipment, and steel have shown significant upward movement, with utilities currently having the highest short-term score of 100 [38][39] - The crowdedness of capital in sectors like electric equipment, steel, and coal has increased, indicating potential volatility risks due to high valuations and sentiment corrections [40][44] - The model indicates a preference for large-cap and value styles, with signals suggesting that these styles may strengthen in the future [49][56]

Quantitative Models and Construction Methods 1. Model Name: Style Rotation Model - **Model Construction Idea**: The model is built from basic style factors such as valuation, market capitalization, volatility, and momentum, gradually constructing a style timing and scoring system[4][9] - **Model Construction Process**: 1. Construct 640 micro features based on 80 basic micro indicators[9] 2. Use common indices as style stock pools to replace the absolute proportion division of style factors, constructing new style returns as labels[4][9] 3. Use a random forest model for style timing and obtain the current score for each style[4][9] 4. Integrate the timing results and scoring results to construct a monthly frequency style rotation model[4][9] - **Model Evaluation**: The model effectively avoids overfitting risks through rolling training of the random forest model and constructs a comprehensive framework from style timing to style scoring and from style scoring to actual investment[9] Model Backtesting Results 1. **Style Rotation Model**: - Annualized Return: 16.18%[10][11] - Volatility: 20.28%[10][11] - Information Ratio (IR): 0.80[10][11] - Win Rate: 59.43%[10][11] - Maximum Drawdown: 25.20%[11] 2. **Market Benchmark (Hedged)**: - Annualized Return: 10.36%[10][11] - Volatility: 10.85%[10][11] - Information Ratio (IR): 0.95[10][11] - Win Rate: 54.72%[10][11] - Maximum Drawdown: 8.53%[11]

Report Industry Investment Rating - The operation rating for CITIC Five Styles - Finance is ★☆☆, indicating a bullish bias but with limited operability in the market [3][4]. Core Viewpoints - As of the week ending October 24, 2025, the weekly returns of Tonglian All A (Shanghai, Shenzhen, Beijing), ChinaBond Composite Bond, and Nanhua Commodity Index were 3.42%, -0.03%, and 0.94% respectively. In the public - fund market, enhanced index strategies led the gains with a weekly return of 3.89%. Neutral strategies had more gains than losses. Among commodities, precious - metal ETFs pulled back, while soybean - meal and non - ferrous - metal ETFs had a slight rebound, and energy - chemical ETFs stabilized [4]. - All CITIC five styles closed up last Friday, with the growth style leading in returns. The style rotation chart showed that the cyclical and consumer styles weakened compared to the previous period, and the growth style had a significant increase in the indicator momentum. In the public - fund pool, financial and cyclical style funds had better excess performance in the past week. The deviation of products from the consumer style increased marginally, and the overall market congestion indicator continued to rise this week, with the growth and financial styles in a historically high - congestion range [4]. - In the neutral strategy, the stock - index basis showed a marginal recovery trend during the week. The IC contract recovered to around 0.5 times the standard deviation above the three - month average. The average premium rates of the spot - index ETFs corresponding to IC and IM were relatively high, in the top 80% quantile range of the past three months [4]. - Among Barra factors, the medium - and long - term momentum factor had a better return performance this week, with a weekly excess return of 1.70%. The residual volatility and ALPHA factors retreated, and the winning rates of the dividend and leverage factors improved. The cross - section rotation speed of factors continued to increase this week, currently in the top 80% quantile range of the past year [4]. - According to the latest scoring results of the style timing model, the growth and financial styles recovered marginally this week, while the cyclical and stable styles declined. The current signal favors the financial style. The return of the style timing strategy last week was 1.45%, and the excess return compared to the benchmark balanced allocation was - 0.98% [4]. Summary by Related Catalogs Fund Market Review Recent Market Returns - The weekly, monthly, quarterly, and semi - annual returns of Tonglian All A (Shanghai, Shenzhen, Beijing), ChinaBond Composite Bond (net), and Nanhua Commodity are presented in a chart [6]. - The maximum drawdowns of the main public - fund strategy indices in the past three months and their weekly returns are also shown in charts [6]. CITIC Style Index - The net - value trends of CITIC style indices (finance, cycle, consumption, growth, stability) from September 24 to October 23, 2025, are presented in a chart [8][9]. - The relative rotation chart of CITIC style indices shows the relative strength and relative - strength momentum of different styles in different time periods (recent week, last week, recent month, recent three months, recent six months, recent year) [10][11]. - The excess - return performance of fund style indices in different time periods is provided in a table [12]. - The fund - style congestion chart shows the congestion levels of cycle, growth, consumption, and finance styles from September 28 to October 26, 2025 [13]. Barra Factors - The style preference of Barra single factors is within the range of 0 - 1, with a higher value indicating a stronger preference. The excess - return performance of Barra single - factor style strategies and the net - value trends of Barra single - factor style excess since this year are presented in charts [14][16][18].

Group 1: Market Sentiment Model Insights - The market sentiment score has slightly increased to 2.2 as of October 24, compared to 1.9 the previous week, indicating a partial recovery in market sentiment [6][8] - The overall market sentiment is showing increased differentiation, with a decline in price-volume consistency, suggesting reduced capital activity [8][12] - The total trading volume for the entire A-share market has significantly decreased compared to the previous week, with a peak trading volume of 1,991.617 billion RMB on October 24 [14][16] Group 2: Sector Performance Insights - As of October 24, the banking, oil and petrochemical, transportation, public utilities, and construction decoration sectors have shown an upward trend in short-term scores [33] - The coal sector currently has the highest short-term score of 93.22, indicating strong short-term performance [33][34] - The model indicates that the market is currently favoring large-cap and value styles, with strong signals for both [33][44] Group 3: Industry Crowding Insights - Recent high price increases in the electronics and power equipment sectors are accompanied by high capital crowding, suggesting potential volatility risks due to valuation and sentiment corrections [36][41] - The average crowding levels are highest in the power equipment, environmental protection, non-ferrous metals, textile and apparel, and coal sectors [37][40] - Low crowding sectors such as non-bank financials, beauty care, media, computing, and food and beverage have shown lower price increases, indicating potential for excess returns if fundamentals improve [36][40]

- **A-share multi-dimensional timing model**: The model evaluates the overall directional judgment of the A-share market using four dimensions: valuation, sentiment, funds, and technical indicators. Each dimension provides daily signals with values of 0, ±1, representing neutral, bullish, or bearish views. Valuation and sentiment dimensions adopt a mean-reversion logic, while funds and technical dimensions use trend-following logic. The final market view is determined by the sum of the scores across all dimensions [9][15][16] - **Style timing model for dividend style**: The model uses three indicators to time the dividend style relative to the CSI Dividend Index and CSI All Share Index. The indicators include relative momentum, 10Y-1Y term spread, and interbank pledged repo transaction volume. Each indicator provides daily signals with values of 0, ±1, representing neutral, bullish, or bearish views. The final view is based on the sum of the scores across all dimensions. When the model favors the dividend style, it fully allocates to the CSI Dividend Index; otherwise, it allocates to the CSI All Share Index [17][21] - **Style timing model for large-cap and small-cap styles**: The model uses momentum difference and turnover ratio difference between the CSI 300 Index and Wind Micro Cap Index to calculate the crowding scores for large-cap and small-cap styles. The model operates in two crowding zones: high crowding and low crowding. In high crowding zones, it uses a small-parameter dual moving average model to address potential style reversals. In low crowding zones, it uses a large-parameter dual moving average model to capture medium- to long-term trends [22][24][26] - **Sector rotation model**: The genetic programming-based sector rotation model selects the top five sectors with the highest multi-factor composite scores from 32 CITIC industry indices for equal-weight allocation. The model updates its factor library quarterly and rebalances weekly. The factors are derived using NSGA-II algorithm, which evaluates factor monotonicity and performance of long positions using |IC| and NDCG@5 metrics. The model combines multiple factors with weak collinearity into sector scores using greedy strategy and variance inflation factor [29][32][33][36] - **China domestic all-weather enhanced portfolio**: The portfolio is constructed using a macro factor risk parity framework, which emphasizes risk diversification across underlying macro risk sources rather than asset classes. The strategy involves three steps: macro quadrant classification and asset selection, quadrant portfolio construction and risk measurement, and risk budgeting to determine quadrant weights. The active allocation is based on macro expectation momentum indicators, which consider buy-side expectation momentum and sell-side expectation deviation momentum [38][41] --- Model Backtesting Results - **A-share multi-dimensional timing model**: Annualized return 24.97%, maximum drawdown -28.46%, Sharpe ratio 1.16, Calmar ratio 0.88, YTD return 37.73%, weekly return 0.00% [14] - **Dividend style timing model**: Annualized return 15.71%, maximum drawdown -25.52%, Sharpe ratio 0.85, Calmar ratio 0.62, YTD return 19.53%, weekly return -3.43% [20] - **Large-cap vs. small-cap style timing model**: Annualized return 26.01%, maximum drawdown -30.86%, Sharpe ratio 1.08, Calmar ratio 0.84, YTD return 64.58%, weekly return -2.22% [27] - **Sector rotation model**: Annualized return 33.33%, annualized volatility 17.89%, Sharpe ratio 1.86, maximum drawdown -19.63%, Calmar ratio 1.70, weekly return 0.14%, YTD return 39.41% [32] - **China domestic all-weather enhanced portfolio**: Annualized return 11.66%, annualized volatility 6.18%, Sharpe ratio 1.89, maximum drawdown -6.30%, Calmar ratio 1.85, weekly return 0.38%, YTD return 10.74% [42]

- The style rotation model is constructed based on the Dongwu quantitative multi-factor system, starting from micro-level stock factors. It selects 80 underlying factors as original features, including valuation, market capitalization, volatility, and momentum, and further constructs 640 micro features. The model replaces the absolute proportion division of style factors with common indices as style stock pools, creating new style returns as labels. A random forest model is trained in a rolling manner to avoid overfitting risks, optimizing features and obtaining style recommendations. The framework integrates style timing, scoring, and actual investment[9][4] - The performance of the style rotation model during the backtesting period (2017/01/01-2025/09/30) shows an annualized return of 16.41%, annualized volatility of 20.43%, IR of 0.80, monthly win rate of 58.49%, and a maximum drawdown of 25.54%. When hedging against the market benchmark, the annualized return is 10.54%, annualized volatility is 10.85%, IR is 0.97, monthly win rate is 55.66%, and the maximum drawdown is 8.79%[10][11] - The style rotation model's latest timing directions for October 2025 are value, large market capitalization, momentum, and low volatility[2][19] - The latest holdings of the style rotation model for October 2025 include indices such as CSI Central Enterprise Dividend (ETF code: 561580.SH), CSI Bank (ETF code: 512700.SH), CSI Film and Television (ETF code: 159855.SZ), CS Battery (ETF code: 159796.SZ), and CSI All Real Estate (ETF code: 512200.SH)[3][19]

Quantitative Models and Construction Methods - **Model Name**: A-Share Multi-Dimensional Timing Model **Model Construction Idea**: The model evaluates the directional judgment of the A-share market using four dimensions: valuation, sentiment, capital, and technical indicators. Valuation and sentiment dimensions adopt a mean-reversion logic, while capital and technical dimensions use trend-following logic. The model combines these dimensions to provide a comprehensive view of market trends [2][9][15]. **Model Construction Process**: 1. The model uses the Wind All A Index as a proxy for the A-share market. 2. Each dimension generates daily signals with values of 0, ±1, representing neutral, bullish, or bearish views. 3. Valuation indicators include equity risk premium (ERP). 4. Sentiment indicators include option put-call ratio, implied volatility, and futures member position ratio. 5. Capital indicators include financing purchase amount. 6. Technical indicators include Bollinger Bands and the difference in the proportion of individual stock trading volume [11][15]. 7. The final multi-dimensional score is calculated as the sum of the scores from the four dimensions, determining the overall market view [9][15]. **Model Evaluation**: The model effectively captures market trends and provides actionable insights for timing decisions [9]. - **Model Name**: Style Timing Model **Model Construction Idea**: The model evaluates timing for dividend and size styles using trend-based indicators and crowding metrics [3][17][22]. **Model Construction Process**: 1. **Dividend Style Timing**: - The model uses three indicators: relative momentum of the CSI Dividend Index vs. CSI All Index, 10Y-1Y term spread, and interbank pledged repo transaction volume. - Each indicator generates daily signals with values of 0, ±1, representing neutral, bullish, or bearish views. - The final score is the sum of the three indicators, determining the overall view on dividend style [17][21]. 2. **Size Style Timing**: - The model uses the crowding degree of small-cap and large-cap styles, calculated based on momentum difference and trading volume ratio between the Wind Micro-Cap Index and CSI 300 Index. - Crowding degree is determined by averaging the top three results of six different window lengths for small-cap and large-cap styles. - High crowding is triggered when small-cap crowding exceeds 90% or large-cap crowding falls below 10%. - In high crowding zones, a small parameter double moving average model is used to capture short-term reversals. In low crowding zones, a large parameter double moving average model is used to follow medium- to long-term trends [22][24][26]. **Model Evaluation**: The model provides effective timing signals for style rotation, especially in different market conditions [22][24]. - **Model Name**: Industry Rotation Model **Model Construction Idea**: The model uses genetic programming to directly extract factors from industry index data, focusing on price-volume and valuation characteristics. It employs a dual-objective genetic programming approach to enhance factor diversity and reduce overfitting [4][29][32]. **Model Construction Process**: 1. The model uses 32 CITIC industry indices as underlying assets. 2. Factors are updated quarterly, and the model rebalances weekly. 3. The dual-objective genetic programming approach evaluates factors using |IC| and NDCG@5 metrics to assess monotonicity and performance of long positions. 4. Factors are combined using a greedy strategy and variance inflation factor to reduce collinearity. 5. The highest-weight factor is constructed as follows: - Perform cross-sectional regression of standardized monthly trading volume against the rolling 4-year percentile of price-to-book ratio (P/B). Take residuals as variable A. - Sum the smallest 9 values of variable A over the past 15 trading days to obtain variable B. - Standardize variable B using z-score, reverse values greater than 2.5, and sum the standardized values over the past 15 trading days [29][33][37]. **Model Evaluation**: The model effectively identifies industry rotation factors with strong monotonicity and performance, while reducing overfitting risks [29][33]. - **Model Name**: China Domestic All-Weather Enhanced Portfolio **Model Construction Idea**: The model adopts a macro factor risk parity framework, emphasizing risk diversification across underlying macro risk sources rather than asset classes. It actively allocates based on macro expectation momentum [5][38][41]. **Model Construction Process**: 1. **Macro Quadrant Division and Asset Selection**: Divide growth and inflation dimensions into four quadrants based on whether they exceed or fall short of expectations. Determine suitable assets for each quadrant using quantitative and qualitative methods. 2. **Quadrant Portfolio Construction and Risk Measurement**: Construct sub-portfolios with equal weights for assets within each quadrant, focusing on downside risk. 3. **Risk Budgeting Model for Quadrant Weights**: Adjust quadrant risk budgets monthly based on "quadrant views" derived from macro expectation momentum indicators, which consider buy-side expectation momentum and sell-side expectation deviation momentum [38][41]. **Model Evaluation**: The model effectively balances macro risks and enhances portfolio performance through active allocation [38][41]. --- Model Backtesting Results - **A-Share Multi-Dimensional Timing Model**: - Annualized Return: 25.23% - Maximum Drawdown: -28.46% - Sharpe Ratio: 1.17 - Calmar Ratio: 0.89 - Year-to-Date (YTD): 40.98% - Last Week's Return: 0.15% [14] - **Style Timing Model**: - **Dividend Style Timing**: - Annualized Return: 16.04% - Maximum Drawdown: -25.52% - Sharpe Ratio: 0.87 - Calmar Ratio: 0.63 - YTD: 21.75% - Last Week's Return: 0.23% [20] - **Size Style Timing**: - Annualized Return: 26.25% - Maximum Drawdown: -30.86% - Sharpe Ratio: 1.09 - Calmar Ratio: 0.85 - YTD: 65.89% - Last Week's Return: 1.07% [27] - **Industry Rotation Model**: - Annualized Return: 32.60% - Annualized Volatility: 17.95% - Sharpe Ratio: 1.82 - Maximum Drawdown: -19.63% - Calmar Ratio: 1.66 - Last Week's Return: 0.27% - YTD: 36.44% [32] - **China Domestic All-Weather Enhanced Portfolio**: - Annualized Return: 11.53% - Annualized Volatility: 6.16% - Sharpe Ratio: 1.87 - Maximum Drawdown: -6.30% - Calmar Ratio: 1.83 - Last Week's Return: 0.66% - YTD: 9.02% [42]

Group 1 - The report focuses on constructing a style timing and rotation model from a micro perspective, utilizing micro data to enhance the strategy system [6][62] - The model is based on four style factors: valuation, market capitalization, volatility, and momentum, using 80 micro indicators to create a scoring system [7][62] - The backtesting period from January 1, 2014, to July 31, 2025, shows an annualized return of 20.90% with a volatility of 26.12% and a maximum drawdown of -40.57% [57][62] Group 2 - The model's out-of-sample performance has been stable since its development in March 2024, with a return of 55.36% for the entire year of 2024, outperforming the market benchmark by 35.72% [57][62] - The report highlights the construction of style labels based on specific broad indices to overcome limitations of using the entire A-share market for style timing [18][20] - The random forest model is selected for predicting the direction of style factors, enhancing the performance of the timing strategy [23][25] Group 3 - The performance metrics for the valuation factor before timing show an annualized return of 7.90% compared to the benchmark's 6.85%, with a maximum drawdown of -60.33% [30][32] - After applying the timing model, the valuation factor's annualized return improves to 15.19%, significantly outperforming the benchmark [38][40] - The momentum factor shows a pre-timing annualized return of 10.11%, which increases to 15.73% post-timing, indicating improved performance [42][47] Group 4 - The volatility factor's pre-timing performance indicates an annualized return of 10.93%, while post-timing performance shows an increase to 15.73% [48][53] - The equal-weighted composite factor, derived from the four style factors, achieves an annualized return of 20.05% with a maximum drawdown of -41.97% [52][55] - The scoring system for the style factors is based on historical prediction accuracy, further refining the composite factor's performance [56][59]

Quantitative Models and Construction Methods - **Model Name**: Style Rotation Model **Model Construction Idea**: The model is built from micro-level stock characteristics, leveraging valuation, market capitalization, volatility, and momentum factors to construct a style timing and scoring system. It integrates micro-level indicators and machine learning techniques to optimize style rotation strategies[4][9] **Model Construction Process**: 1. Select 80 base factors as original features based on the Dongwu multi-factor system[9] 2. Construct 640 micro-level features from these base factors[4][9] 3. Replace absolute proportion division of style factors with common indices as style stock pools to create new style returns as labels[4][9] 4. Use rolling training with a Random Forest model to avoid overfitting risks, optimize feature selection, and generate style recommendations[4][9] 5. Develop a framework from style timing to scoring, and from scoring to actual investment decisions[9] **Model Evaluation**: The model effectively avoids overfitting risks and provides a comprehensive framework for style rotation strategies[9] Model Backtesting Results - **Style Rotation Model**: - Annualized Return: 16.66%[10][11] - Annualized Volatility: 19.57%[10][11] - Information Ratio (IR): 0.85[10][11] - Monthly Win Rate: 56.31%[10][11] - Maximum Drawdown: -29.34%[11] - Excess Return (vs Benchmark): 11.40%[10][11] - Excess Volatility (vs Benchmark): 13.04%[10][11] - Excess IR (vs Benchmark): 0.87[10][11] - Excess Monthly Win Rate (vs Benchmark): 57.28%[10][11] - Excess Maximum Drawdown (vs Benchmark): -9.73%[11] Quantitative Factors and Construction Methods - **Factor Name**: Valuation, Market Capitalization, Volatility, Momentum **Factor Construction Idea**: These factors are derived from micro-level stock characteristics and are used to construct style timing and scoring systems[4][9] **Factor Construction Process**: 1. Extract micro-level features from base factors[4][9] 2. Use these features to create style returns as labels for machine learning models[4][9] 3. Apply Random Forest models to optimize factor selection and timing[4][9] **Factor Evaluation**: These factors are foundational to the style rotation model and contribute to its effectiveness in timing and scoring[4][9] Factor Backtesting Results - **Valuation Factor**: Monthly Returns (2025/01-2025/05): -2.00%, 0.00%, 2.00%, 4.00%, 6.00%[13][20] - **Market Capitalization Factor**: Monthly Returns (2025/01-2025/05): -4.00%, -2.00%, 0.00%, 2.00%, 4.00%[13][20] - **Volatility Factor**: Monthly Returns (2025/01-2025/05): -6.00%, -4.00%, -2.00%, 0.00%, 2.00%[13][20] - **Momentum Factor**: Monthly Returns (2025/01-2025/05): -8.00%, -6.00%, -4.00%, -2.00%, 0.00%[13][20]