Quantitative Models and Factor Construction Quantitative Models and Construction Methods - **Model Name**: Market Cap Segmented Linear Regression Model **Construction Idea**: Adjust the weights of factor regressions based on market cap segmentation to address the performance differences of factors across different market cap groups [7][10][12] **Construction Process**: 1. Factors are divided into five categories: Dividend, ROE_SUE, Daily Volume-Price, High-Frequency Volume-Price, and a final composite factor [7][10] 2. Use OLS regression with IC or ICIR weighting to combine sub-factors into composite factors [7] 3. Apply KMedian clustering on the log of market cap to divide stocks into 11 groups [7] 4. Assign weights to each group using the formula: $ w_{i}=w_{base}+(1-w_{base})*|i-I|/n $ where $w_{base}$ is the minimum weight (set to 0.9, 0.5, or 0), $n$ is the number of groups, and $I$ is the group with the highest weight [7] 5. Train 11 models with different weight assignments and evaluate the composite factor's IC, RankMAE, long-short returns, and long-only returns [7] **Evaluation**: This model improves factor performance in specific market cap segments, particularly for small-cap stocks, but extreme weighting can increase volatility [7][12] - **Model Name**: Market Cap Weighted Composite Factor Model **Construction Idea**: Reweight composite factors based on market cap distribution to enhance factor performance in specific indices [48][49][65] **Construction Process**: 1. Use market cap weights from benchmark indices (e.g., CSI 300, CSI 500, CSI 1000) to reweight composite factors [48] 2. Construct enhanced portfolios with weekly rebalancing and constraints on individual stock weights, industry weights, and turnover [48] **Evaluation**: Significant performance improvement in CSI 300 and CSI 500 indices, with annualized excess returns increasing by over 1% in some cases. However, the method is less effective for CSI 1000 [49][65][79] - **Model Name**: Market Cap Weighted Cross-Composite Factor Model **Construction Idea**: Match factor weights to the market cap group of each stock to reduce parameter sensitivity [80][81] **Construction Process**: 1. Assign factor values based on the stock's market cap group: $ F_{i}=F_{l_{i}}\;\;i\in I $ where $i$ belongs to market cap group $I$ [80] 2. Evaluate single-factor performance and construct enhanced portfolios for different indices [81][85] **Evaluation**: Performance improvement is observed in CSI 300 and CSI 500 indices, but the method is less effective for CSI 1000. Parameter sensitivity is reduced compared to other methods [85][92][96] - **Model Name**: Multi-Style Factor Weighted Composite Factor Model **Construction Idea**: Incorporate style factors (e.g., value-growth, industry) into the weighting process to address factor performance differences across styles [98][99] **Construction Process**: 1. Cluster stocks based on style factors using Manhattan distance [98] 2. Construct 11 composite factor models centered on each style cluster [98] 3. Use cross-composite and component-composite methods to evaluate performance in enhanced portfolios [100][101] **Evaluation**: Performance improvement is limited compared to market cap-based methods. Cross-composite weighting shows better results than component-composite weighting in some cases [101][115][132] Backtest Results of Models - **Market Cap Segmented Linear Regression Model**: - IC: 0.057 (all-market), 0.037 (CSI 300), 0.040 (CSI 500), 0.052 (CSI 1000), 0.060 (small-cap) [7][81][84] - RankMAE: 1.090 (all-market), 1.119 (CSI 300), 1.111 (CSI 500), 1.106 (CSI 1000), 1.092 (small-cap) [7][81][84] - Long-Short Returns: 1.07% (all-market), 0.38% (CSI 300), 0.49% (CSI 500), 0.92% (CSI 1000), 1.19% (small-cap) [7][81][84] - **Market Cap Weighted Composite Factor Model**: - CSI 300: Annualized Return 8.21%, IR 0.966, Max Drawdown 15.67% (base_w=0) [49] - CSI 500: Annualized Return 14.64%, IR 1.385, Max Drawdown 12.60% (base_w=0.5) [59] - CSI 1000: Annualized Return 18.95%, IR 1.585, Max Drawdown 16.59% (equal weight) [70] - **Market Cap Weighted Cross-Composite Factor Model**: - CSI 300: Annualized Return 7.36%, IR 0.901, Max Drawdown 16.33% (base_w=0) [85] - CSI 500: Annualized Return 15.06%, IR 1.409, Max Drawdown 13.14% (base_w=0.5) [92] - CSI 1000: Annualized Return 18.95%, IR 1.585, Max Drawdown 16.59% (equal weight) [92] - **Multi-Style Factor Weighted Composite Factor Model**: - CSI 300: Annualized Return 7.24%, IR 0.926, Max Drawdown 16.32% (base_w=0.9, component-composite) [103] - CSI 500: Annualized Return 14.17%, IR 1.377, Max Drawdown 12.65% (base_w=0, cross-composite) [115] - CSI 1000: Annualized Return 18.63%, IR 1.570, Max Drawdown 16.47% (base_w=0, component-composite) [132]

Core Insights - The report constructs a "limited attention" factor that represents the degree of retail investor attention on individual stocks, using indicators such as abnormal turnover, abnormal trading volume, extreme returns, and whether the stock has appeared on the "Dragon and Tiger List" [3][6][56] - The limited attention factor can indicate a "money-making effect," showing a high success rate for timing signals, with an overall monthly success rate exceeding 70% when applied to the CSI 300 index [3][25][56] - The factor performs better in smaller stock pools, indicating that retail investors' "herding behavior" can lead to significant price fluctuations in these stocks [3][56] Limited Attention Factor Construction - The construction of the limited attention factor is based on the premise that retail investors prioritize stocks that attract their attention, leading to a focus on high turnover, high trading volume, and extreme returns [6][8] - The factor is constructed using both linear combination and random forest methods, with the latter showing better predictive power for returns [13][14][19] Performance of the Limited Attention Factor - The performance of the limited attention factor is evaluated through its information coefficient (IC) and the average change in shareholder accounts, indicating a clear relationship between higher attention levels and increased retail investor interest [18][19] - The factor's IC and monthly long-short returns are notably higher in the CSI 1000 index, aligning with the logic that smaller stocks are more susceptible to retail investor behavior [22][24] Money-Making Effect Indication - The report highlights that the limited attention factor's IC can reflect the strength of the market's "money-making effect," particularly during the period from 2019 to mid-2021 when retail investors showed strong interest in high attention stocks [25][27][32] - The correlation between the limited attention factor's IC and industry trend consistency indicates that higher IC values suggest better market performance [29][30] Application of the Limited Attention Factor - The factor can be directly used in index enhancement strategies, either by adding it to existing models or by excluding stocks with high limited attention [35][37] - The report finds that directly adding the limited attention factor improves performance in the CSI 300 and CSI 500 indices, while both methods fail to enhance returns in the CSI 1000 index [40][57] Performance of Other Factors in Limited Attention Domains - The report identifies that price-volume factors, particularly low volatility, low liquidity, and long-term momentum, perform significantly better in the limited attention domain, while profitability, valuation, and dividend factors show decreased effectiveness [41][42][43] - The analysis of different stock pools reveals that the performance of factors varies significantly between limited attention and non-limited attention stocks, with growth and price-volume factors being more effective in the former [41][44]

Quantitative Models and Construction Methods Model Name: Black Swan Index - **Construction Idea**: Measure the extremity of market transactions based on the deviation of style factor returns[24][25] - **Construction Process**: 1. Calculate the daily return deviation of style factors: $$ \sigma_{s,t}=\frac{\bar{r}_{s,t}-\bar{r}_{s}}{\sigma_{s}} $$ where $\bar{r}_{s,t}$ is the daily return of style factor $s$ on day $t$, $\bar{r}_{s}$ is the average daily return of style factor $s$ over the entire sample period, and $\sigma_{s}$ is the standard deviation of daily returns of style factor $s$ over the entire sample period[25] 2. Calculate the Black Swan Index: $$ BlackSwan_{t}=\frac{1}{N}\times\sum_{s\in S}\left|\sigma_{s,t}\right| $$ where $BlackSwan_{t}$ is the Black Swan Index on day $t$, $S$ is the set of all style factors, and $N$ is the number of style factors[25] - **Evaluation**: The Black Swan Index effectively captures the extremity of market transactions, indicating higher probabilities of extreme tail risks[24][25] Model Name: Heuristic Style Classification for Cognitive Risk Control - **Construction Idea**: Address the discrepancy between individual and collective cognition in style classification to control cognitive risk[80][81] - **Construction Process**: 1. Calculate the value and growth factors for each stock based on predefined metrics[85] 2. Construct value and growth portfolios by selecting the top 10% and bottom 10% stocks based on factor scores[82] 3. Perform time-series regression to classify stocks into value, growth, or balanced styles: $$ r_{t,t}\sim\beta_{\mathit{Value}}\cdot r_{\mathit{Value},t}+\beta_{\mathit{Growth}}\cdot r_{\mathit{Growth},t}+\varepsilon_{t} $$ subject to $0\leq\beta_{\mathit{Value}}\leq1$, $0\leq\beta_{\mathit{Growth}}\leq1$, and $\beta_{\mathit{Value}}+\beta_{\mathit{Growth}}=1$[97] 4. Use weighted least squares (WLS) to estimate regression coefficients based on the most differentiated trading days[98] - **Evaluation**: The heuristic style classification method captures market consensus more accurately than traditional factor scoring methods, reducing cognitive risk[80][81] Model Name: Louvain Community Detection for Hidden Risk Control - **Construction Idea**: Cluster stocks based on excess return correlations to identify hidden risks[116][117] - **Construction Process**: 1. Calculate weighted correlation of excess returns between stocks: $$ Corr_{w}(X,Y)=\frac{Cov_{w}(X,Y)}{\sigma_{w,X}\cdot\sigma_{w,Y}}=\frac{\sum_{i=1}^{n}w_{i}(x_{i}-\overline{X_{w}})(y_{i}-\overline{Y_{w}})}{\sqrt{\sum_{i=1}^{n}w_{i}(x_{i}-\overline{X_{w}})^{2}}\cdot\sqrt{\sum_{i=1}^{n}w_{i}(y_{i}-\overline{Y_{w}})^{2}}} $$ where $w_{i}$ is the weight for day $i$, reflecting market volatility[118] 2. Use Louvain algorithm to cluster stocks based on weighted correlation matrix[117] 3. Ensure clusters have at least 20 stocks and remove clusters with fewer stocks[121] - **Evaluation**: The Louvain community detection method effectively identifies hidden risks by clustering stocks with similar return patterns, which traditional risk models may overlook[116][117] Model Name: Dynamic Style Factor Control - **Construction Idea**: Control style factors dynamically based on their volatility clustering effect[128][129] - **Construction Process**: 1. Identify style factors with high volatility or significant volatility increase: $$ \text{High volatility: Rolling 3-month volatility in top 3} $$ $$ \text{Volatility increase: Rolling 3-month volatility > historical mean + 1 standard deviation} $$ 2. Set the exposure of these style factors to zero in the portfolio[136] - **Evaluation**: Dynamic style factor control captures major market risks without significantly affecting portfolio returns, leveraging the predictability of volatility clustering[128][129] Model Name: Adaptive Stock Deviation Control under Target Tracking Error - **Construction Idea**: Adjust stock deviation based on tracking error to control portfolio risk[146][147] - **Construction Process**: 1. Calculate rolling 3-month tracking error for different stock deviation levels[153] 2. Set the maximum stock deviation that keeps tracking error within the target range[153] - **Evaluation**: Adaptive stock deviation control effectively reduces tracking error during high market volatility, maintaining portfolio stability[146][147] Model Backtest Results Traditional CSI 500 Enhanced Index - **Annualized Excess Return**: 18.77%[5][162] - **Maximum Drawdown**: 9.68%[5][162] - **Information Ratio (IR)**: 3.56[5][162] - **Return-to-Drawdown Ratio**: 1.94[5][162] - **Annualized Tracking Error**: 4.88%[5][162] CSI 500 Enhanced Index with Full-Process Risk Control - **Annualized Excess Return**: 16.51%[5][169] - **Maximum Drawdown**: 4.90%[5][169] - **Information Ratio (IR)**: 3.94[5][169] - **Return-to-Drawdown Ratio**: 3.37[5][169] - **Annualized Tracking Error**: 3.98%[5][169]

一、本周指数增强组合表现 沪深300指数增强组合本周超额收益0.77%，本年超额收益3.14%。 中证500指数增强组合本周超额收益1.14%，本年超额收益3.91%。 中证1000指数增强组合本周超额收益0.79%，本年超额收益5.21%。 二、本周选股因子表现跟踪 沪深300成分股中3个月盈利上下调、标准化预期外盈利、单季净利同比增速 等因子表现较好。 中证500成分股中预期PEG、BP、标准化预期外收入等因子表现较好。 中证1000成分股中预期净利润环比、预期PEG、单季超预期幅度等因子表现 较好。 公募基金重仓股中一年动量、3个月盈利上下调、单季净利同比增速等因子 表现较好。 三、本周公募基金指数增强产品表现跟踪 沪深300指数增强产品本周超额收益最高2.02%，最低-0.56%，中位数 0.45%。 中证500指数增强产品本周超额收益最高1.36%，最低-0.28%，中位数 0.59%。 中证1000指数增强产品本周超额收益最高1.44%，最低-0.17%，中位数 0.78%。 主 要 结 论 一 国信金工指数增强组合表现跟踪 国信金工指数增强组合的构建流程主要包括收益预测、风险控制和组合优化三块， ...

Group 1 - The core viewpoint of the article is to track the performance of index-enhanced portfolios and stock selection factors across different indices, highlighting their excess returns and the effectiveness of various stock selection factors [1][2][14]. Group 2 - The performance of the CSI 300 index-enhanced portfolio showed an excess return of -1.25% for the week and 1.61% year-to-date [1][2]. - The performance of the CSI 500 index-enhanced portfolio indicated an excess return of -1.53% for the week and 2.17% year-to-date [1][2]. - The performance of the CSI 1000 index-enhanced portfolio recorded an excess return of -0.88% for the week and 3.15% year-to-date [1][2]. Group 3 - In the CSI 300 sample space, factors such as non-liquidity shock, three-month turnover, and one-month turnover performed well recently [4][5]. - In the CSI 500 sample space, factors like expected net profit month-on-month, non-liquidity shock, and three-month earnings adjustments showed strong performance [6][8]. - In the CSI 1000 sample space, factors including three-month institutional coverage and turnover metrics performed well [9][10]. Group 4 - The public fund index-enhanced products for the CSI 300 had a maximum excess return of 1.04% and a minimum of -2.85% for the week, with a median of -0.53% [16]. - The public fund index-enhanced products for the CSI 500 had a maximum excess return of 0.86% and a minimum of -1.80% for the week, with a median of -0.62% [18]. - The public fund index-enhanced products for the CSI 1000 had a maximum excess return of 0.86% and a minimum of -1.80% for the week, with a median of -0.62% [18]. Group 5 - The total number of public fund index-enhanced products includes 67 for the CSI 300 with a total scale of 81.5 billion, 68 for the CSI 500 with a total scale of 49.3 billion, and 46 for the CSI 1000 with a total scale of 16.9 billion [15].

Group 1 - The core viewpoint of the article is to track the performance of index-enhanced portfolios and stock selection factors across different indices, highlighting their excess returns and factor effectiveness [1][2][3]. Group 2 - The performance of the CSI 300 index-enhanced portfolio showed an excess return of 0.15% for the week and 0.96% for the year [1][2]. - The performance of the CSI 500 index-enhanced portfolio indicated an excess return of -0.12% for the week and 1.77% for the year [1][2]. - The performance of the CSI 1000 index-enhanced portfolio reflected an excess return of -0.62% for the week and -0.14% for the year [1][2]. Group 3 - In the CSI 300 component stocks, factors such as dividend yield, DELTAROA, and three-month institutional coverage performed well [1]. - In the CSI 500 component stocks, factors like one-year momentum, DELTAROA, and standardized expected excess income showed strong performance [1]. - In the CSI 1000 component stocks, factors including quarterly net profit year-on-year growth, DELTAROA, and quarterly revenue year-on-year growth were effective [1]. Group 4 - The public fund index-enhanced products for the CSI 300 had a maximum excess return of 1.34% and a minimum of -0.63% for the week, with a median of 0.12% [1][17]. - The public fund index-enhanced products for the CSI 500 had a maximum excess return of 0.97% and a minimum of -0.82% for the week, with a median of 0.02% [1][19]. - The public fund index-enhanced products for the CSI 1000 had a maximum excess return of 1.24% and a minimum of -1.00% for the week, with a median of -0.01% [1][21].

Group 1 - The core viewpoint of the article highlights the performance of various index-enhanced portfolios, indicating that the CSI 300 and CSI 500 index-enhanced portfolios achieved positive excess returns, while the CSI 1000 index-enhanced portfolio experienced a slight decline in excess returns this week [1][2][18]. Group 2 - The CSI 300 index-enhanced portfolio recorded an excess return of 0.44% this week and 0.80% year-to-date [1][2]. - The CSI 500 index-enhanced portfolio also achieved an excess return of 0.44% this week and 1.81% year-to-date [1][2]. - The CSI 1000 index-enhanced portfolio saw a decrease of 0.13% in excess return this week, with a year-to-date excess return of 0.50% [1][2]. Group 3 - In the CSI 300 component stocks, factors such as three-month turnover, dividend yield, and one-month turnover performed well [5]. - For the CSI 500 component stocks, factors like executive compensation, expected net profit month-on-month, and dividend yield showed strong performance [6]. - In the CSI 1000 component stocks, factors such as expected PEG, SPTTM, and one-month volatility performed well [10]. Group 4 - The public fund index-enhanced products for the CSI 300 had a maximum excess return of 1.67%, a minimum of -2.70%, and a median of 0.11% this week [18]. - The CSI 500 index-enhanced products had a maximum excess return of 1.55%, a minimum of -0.45%, and a median of 0.38% this week [19]. - The CSI 1000 index-enhanced products recorded a maximum excess return of 1.59%, a minimum of -0.87%, and a median of 0.30% this week [21].