广 广发证券 广发证券2025年秋季策略会-余工论坛 智慧量化·未来投资 16:00 周飞鹏 广发证券金融工程资深分析师 基于多因子加权的ETF轮动策略 16:00 16:30 张钰东 广发证券金融工程资深分析师 如何利用聪明钱改进分析师预期 16:30 17:00 王小康 广发证券金融工程资深分析师 (98月27日 13:30-17:00 0 浦东香格里拉大酒店 浦江楼2楼北京厅 会议流程 多角度定量刻画指数拥挤度 15:00 AI复盘之精选策略组合 13:30 14:00 安宁宁 广发证券金融工程首席分析师 面向通用模型的时序增强学习 14:00 14:30 陈原文 广发证券金融工程联席首席分析师 可转债指数择时的三个视角 14:30 15:00 张 超 广发证券金融工程资深分析师 15:30 李 豪 广发证券金融工程资深分析师 龙头扩散效应行业轮动 15:30 - 欢迎 扫 码 报 名 - ...

Core Viewpoint - The recent market performance shows a significant increase in the ChiNext and Sci-Tech 50 indices, while large-cap value stocks have declined, indicating a shift in investor sentiment towards growth sectors [1][2]. Market Performance - In the last five trading days, the Sci-Tech 50 index rose by 5.53%, the ChiNext index increased by 8.48%, while the large-cap value index fell by 0.76%. The large-cap growth index rose by 3.63%, and the Shanghai 50 index increased by 1.57%. Small-cap stocks represented by the CSI 2000 index rose by 3.86% [1]. - The communication and electronics sectors performed well, while the banking and steel sectors lagged behind [1]. Risk Premium Analysis - The risk premium, measured as the difference between the inverse of the static PE of the CSI All Share Index and the yield of ten-year government bonds, has reached historical extremes. As of October 28, 2022, the risk premium was at 4.08%, indicating a potential market rebound [1]. - The risk premium has exceeded 4% for the fifth time since 2016, with the latest reading on January 19, 2024, at 4.11% [1]. Valuation Levels - As of August 15, 2025, the CSI All Share Index's TTM PE is at the 72nd percentile, with the Shanghai 50 and CSI 300 at 69% and 63%, respectively. The ChiNext index is at a relatively low valuation level of approximately 33% [2]. - The long-term view of the Deep 100 index suggests a cyclical pattern of bear and bull markets every three years, with the current adjustment phase starting in Q1 2021 showing sufficient time and space for a potential upward cycle [2]. Fund Flow and Trading Activity - In the last five trading days, there was an outflow of 10.4 billion yuan from ETFs, while margin financing increased by approximately 41.8 billion yuan. The average daily trading volume across both markets was 20,767 billion yuan [3]. AI and Trend Observation - The use of convolutional neural networks (CNN) for modeling price and volume data has been explored, with the latest focus on mapping learned features to industry themes, particularly in the communication sector [8].

Core Viewpoint - The report focuses on constructing stock price jump-related factors based on Level 2 data to detect the effects of volatility, jump amplitude, and jump trading activity on stock selection [1][4]. Group 1: Research Background - The report aims to utilize Level 2 data for in-depth analysis to uncover hidden market patterns, which may include stock price trends and short-term trading signals [3][4]. - Previous studies have introduced jump-diffusion models and derived various jump-related factors for empirical testing [3][4]. Group 2: Level 1 and Level 2 Market Data - Level 1 data includes basic trading information such as highest price, lowest price, opening price, closing price, trading volume, and trading amount, while Level 2 data provides more detailed information, including tick data and order book depth [5][6]. Group 3: Jump Factor Research - The jump factors are categorized into jump volatility, cumulative jump values, and trading volume ratios, with further refinements based on the direction and size of jumps [18][21]. - The report discusses the construction of jump volatility factors and cumulative jump value factors, which consider the positive and negative directions of price changes [19][23]. Group 4: Empirical Backtesting - The backtesting period is set from January 1, 2020, to July 18, 2025, with a focus on the performance of various factors under different trading frequencies [27]. - The RRJV factor shows a RANK_IC of 8.18% and an annualized return of 27.8% during the backtesting period [28][29]. - The JSR2_drop factor achieves a RANK_IC of 9.77% with an annualized return of 22.1% [30]. Group 5: Performance Analysis - The report indicates that the performance of jump-related factors varies with trading frequency, with some factors showing improved returns under weekly trading strategies [32][36]. - The RRJV factor's backtesting results indicate a significant increase in long-term returns, reaching 28.2% under weekly trading conditions [36].

Market Performance - The recent 5 trading days saw the Sci-Tech 50 Index increase by 0.65%, the ChiNext Index by 0.49%, the large-cap value by 1.63%, the large-cap growth by 1.17%, the SSE 50 by 1.27%, and the small-cap represented by the CSI 2000 by 2.74% [1] - The sectors of defense, military, and non-ferrous metals performed well, while pharmaceuticals, biotechnology, and computers lagged behind [1] Risk Premium Analysis - The static PE of the CSI All Index minus the yield of 10-year government bonds indicates a risk premium, which reached 4.17% on April 26, 2022, and 4.08% on October 28, 2022, showing a market rebound [1] - As of January 19, 2024, the risk premium indicator was at 4.11%, marking the fifth time since 2016 it exceeded 4% [1] - The indicator as of August 8, 2025, was at 3.39%, with the two-standard deviation boundary at 4.77% [1] Valuation Levels - As of August 8, 2025, the CSI All Index's PE TTM percentile was at 68%, with the SSE 50 and CSI 300 at 69% and 61% respectively, while the ChiNext Index was close to 25% [2] - The long-term view of the Deep 100 Index shows a technical pattern of bear markets every three years followed by bull markets, with the current adjustment starting in Q1 2021 being substantial [2] Fund Flow and Trading Activity - In the last 5 trading days, ETF inflows amounted to 18.5 billion yuan, and the margin trading increased by approximately 27.8 billion yuan, with an average daily trading volume of 1.6748 trillion yuan [3] Neural Network Trend Observation - A convolutional neural network was utilized to model price and volume data, mapping learned features to industry themes, with a focus on semiconductor materials among the latest configurations [9]

Group 1 - The core viewpoint of the article is the rapid development of ETFs in the A-share market since 2019, with the total number of ETFs increasing from 198 in Q4 2018 to 1209 by Q2 2025, and the total scale rising from 0.51 trillion yuan to 4.31 trillion yuan during the same period [1][9][11] - As of Q2 2025, stock-type ETFs account for approximately 75% of the total scale, while cross-border ETFs, currency-type ETFs, bond-type ETFs, and commodity-type ETFs collectively account for about 25% [11][12] Group 2 - The article discusses the construction of an all-weather diversified ETF portfolio based on various methods, including A-share market asset allocation ETFs, overseas equity index QDII-ETFs, relative return ETFs, and absolute return ETFs [2][13] - The methodology for constructing the A-share market asset allocation ETF portfolio and overseas equity index QDII-ETF portfolio involves quantitative scoring based on macro and technical perspectives [4][19] Group 3 - The performance of the all-weather diversified ETF portfolio shows an annualized return of 9.22% with a maximum drawdown of 3.64% and an annualized volatility of 3.85% from December 31, 2016, to June 30, 2025 [6][40] - The portfolio consistently achieved positive absolute returns across different years within the backtesting period, indicating a balanced source of returns [6][40] Group 4 - The construction method for the A-share market asset allocation ETF portfolio includes selecting large-cap indices such as CSI 300 and CSI 500, and bond indices like 1-5 year national development bonds [22][23] - The strategic allocation models used include fixed ratio, volatility control, and risk parity models, with monthly adjustments based on macro and technical indicators [35][36] Group 5 - The QDII-ETF portfolio construction method focuses on overseas equity indices, including the Hang Seng Index, S&P 500, and Nikkei 225, with macro and technical indicators influencing the scoring and weighting [43][44] - The QDII-ETF portfolio achieved an annualized return of 17.24% with a maximum drawdown of 17.20% during the backtesting period [54][55] Group 6 - The relative return ETF portfolio is constructed based on six dimensions, including historical fundamentals and capital flow, to implement an index rotation strategy [57][60] - The portfolio that incorporates index crowding indicators outperformed the standard relative return ETF portfolio, achieving an annualized return of 16.59% compared to 14.04% [79] Group 7 - The absolute return ETF portfolio is designed to focus on absolute returns by selecting indices with stable fundamentals and significant dividend yields, while minimizing exposure to market volatility [83]

Core Viewpoint - The article discusses the construction and testing of index congestion indicators to enhance the performance of ETF portfolios by removing ETFs corresponding to highly congested indices, thereby reducing the impact of market reversals on the ETF portfolio [1][2][3]. Group 1: Index Congestion Indicators - The construction of congestion indicators is based on six dimensions: trading volume, volatility level, financing balance, financing increment, fund holdings, and capital flow [2][23]. - The effectiveness of these congestion indicators is tested by comparing the performance of a congestion index portfolio against the average performance of sample equity indices [25][55]. - The overall correlation between different congestion indicators is low, indicating that a multi-indicator congestion index portfolio performs more stably [2][55]. Group 2: ETF Portfolio Construction - The article outlines the process of constructing relative return index portfolios by excluding indices with two or more congestion indicators from the top-scoring indices [3][63]. - Backtesting results show that the constructed portfolios outperform those that do not consider index congestion, with a cumulative return of 355.05% from December 31, 2016, to June 30, 2025 [12][64]. - The annualized return of the relative return index portfolio combined with index congestion is 19.75%, compared to 17.67% for the standard relative return portfolio [64][66]. Group 3: A-share Market ETF Development - Since Q4 2018, the number of equity ETFs in the A-share market has increased from 133 to 972 by Q2 2025, with total assets rising from 0.27 trillion yuan to 3.03 trillion yuan [7]. Group 4: Backtesting Results - The cumulative return of the relative return ETF portfolio from December 31, 2016, to June 30, 2025, is 201.79%, significantly higher than the benchmark portfolio's return of 33.38% [12][70]. - The performance of the relative return ETF portfolio shows significant excess returns in most years during the backtesting period [12][66]. Group 5: Individual Congestion Indicators - The article details the performance of individual congestion indicators, such as trading volume and beta, showing that portfolios based on these indicators generally underperform compared to sample equity indices [26][33][38]. - For example, the trading volume congestion index portfolio had a cumulative return of -11.66% compared to 41.03% for the sample equity index portfolio [26][33]. Group 6: Multi-Indicator Combination - A multi-indicator congestion index portfolio is constructed by combining the six different congestion indicators, which shows a low average correlation among them [55][58]. - Backtesting results indicate that multi-indicator portfolios generally underperform compared to sample equity indices, particularly those with two or more congestion indicators [59][70].

Market Performance - The recent five trading days saw the Sci-Tech 50 Index decline by 1.65%, the ChiNext Index by 0.74%, the large-cap value index by 1.27%, the large-cap growth index by 2.58%, the SSE 50 by 1.48%, and the CSI 2000 representing small caps by 0.19% [1] - The pharmaceutical and communication sectors performed well, while coal and non-ferrous metals lagged [1] Risk Premium Analysis - The risk premium, defined as the inverse of the static PE of the CSI All Index (EP) minus the yield of ten-year government bonds, indicates that the implied returns of equity and bond assets are at historically high levels, reaching 4.17% on April 26, 2022, and 4.08% on October 28, 2022 [1] - As of January 19, 2024, the indicator was at 4.11%, marking the fifth occurrence since 2016 of exceeding 4% [1] - The latest figure as of August 1, 2025, is 3.48%, with the two-standard-deviation boundary set at 4.76% [1] Valuation Levels - As of August 1, 2025, the CSI All Index's TTM PE is at the 64th percentile, with the SSE 50 and CSI 300 at 66% and 58% respectively, while the ChiNext Index is close to 25% [2] - The CSI 500 and CSI 1000 are at 46% and 37% respectively, indicating that the ChiNext Index's valuation is relatively low compared to historical levels [2] Long-term Market Trends - The technical analysis of the Deep 100 Index shows a pattern of bear markets every three years, followed by bull markets, with previous declines ranging from 40% to 45% [2] - The current adjustment cycle began in the first quarter of 2021, suggesting a potential for upward movement from the bottom [2] Fund Flow and Trading Activity - In the last five trading days, ETF funds experienced an outflow of 13.1 billion yuan, while margin financing increased by approximately 42.6 billion yuan [2] - The average daily trading volume across both markets was 1.7848 trillion yuan [2] AI and Machine Learning Applications - A convolutional neural network (CNN) was utilized to model price and volume data, mapping learned features to industry themes, with a focus on semiconductor materials [2][7] ETF Indexes - Various ETF indexes related to semiconductor materials and innovation were listed, including the SSE Sci-Tech Semiconductor Materials Equipment Theme Index and the CSI Semiconductor Industry Index, all dated August 1, 2025 [8]

Core Viewpoint - Temporal Data Augmentation is increasingly recognized as a technique to enhance the generalization ability and robustness of quantitative models in finance, addressing the challenge of homogeneous data sources among investors [1][4][5]. Group 1: Temporal Data Augmentation - Temporal Data Augmentation involves various strategies such as shifting, scaling, perturbation, cropping, and synthesis to create a richer training sample space without introducing additional information [1][4]. - This technique is applicable not only to traditional machine learning models but also seamlessly integrates into deep learning architectures and reinforcement learning systems, expanding the expressiveness and adaptability of quantitative strategies [1][4]. Group 2: Application Methodology - The study uses GRU as a representative deep learning model to explore whether Temporal Data Augmentation can improve performance while keeping the original input data, network, loss function, and hyperparameter settings consistent [1][58]. - Two training modes are discussed: one with a fixed probability p for data augmentation and another with a linearly decaying probability p throughout the training process [2][63]. Group 3: Empirical Analysis - In the fixed probability p training mode, no significant improvement in factor performance was observed; however, in the linearly decaying probability p mode, various data augmentation factors showed improvements in RankIC and annualized returns [2][67]. - Specifically, the RankIC mean increased by 1.2%, and the annualized returns for long and short positions improved by 2.81% and 7.65%, respectively, when combining data augmentation factors with original data factors [2][75]. Group 4: Data Augmentation Techniques - The study identifies eight different temporal data augmentation techniques, including jittering, scaling, rotation, permutation, magnitude warping, time warping, window slicing, and window warping, and compares their performance against the original data [58][67]. - Among these techniques, jittering and scaling showed the highest correlation with the original data, indicating minimal disruption to the temporal information [59]. Group 5: Performance Metrics - The performance metrics for the various data augmentation methods under fixed probability p indicate that jittering and scaling achieved the highest RankIC win rates, while rotation and time warping resulted in significant information loss [68]. - In the linearly decaying probability p mode, jittering demonstrated the most substantial performance improvement, with a RankIC mean of 13.30% and an annualized return of 55.35% [75].

Core Viewpoint - The report emphasizes the importance of analyst prediction factors, including analyst coverage, growth predictions, and adjustments, and how their performance has been affected by market changes and trading structures [1][4][55]. Analyst Prediction Factors Introduction - Fundamental factors like analyst predictions have been favored by quantitative investors due to their logical basis, but they have shown significant cyclical fluctuations in recent years [4][5]. - The report focuses on enhancing the performance of prediction factors using various data, including price and volume [4]. Analyst Coverage Factor Testing and Improvement - As of the end of 2024, analysts covered 3,142 stocks, representing a coverage rate of 58.30% of the total A-share market [7]. - In major stock pools, coverage rates are high, with 100% in the CSI 300 and 93.31% in the CSI 500 [9]. - The adjusted coverage factor showed a significant improvement, with an annualized return of 12.62% and a Sharpe ratio of 1.61 after controlling for trading volume and market attention [20][55]. Smart Money and Analyst Behavior - The report explores the impact of "smart money" on analyst predictions, indicating that stocks with lower smart money participation before report releases tend to yield higher excess returns when analysts are optimistic [30][33]. - A new grouping method based on smart money indicators significantly improved the monotonicity of excess returns, with the top group showing a 0.60% return and the bottom group showing -1.02% over 20 days [37][55]. Analyst Earnings Prediction Factor Testing and Improvement - The performance of growth and adjustment prediction factors has declined due to changes in market pricing logic and trading behaviors [26][29]. - After adjustments, the improved prediction factors showed notable increases in performance metrics, with the adjusted ROE growth factor in the CSI 300 achieving an IC mean of 4.90% and an annualized return of 9.62% [40][56]. Summary - The report concludes that adjusting analyst prediction factors using smart money indicators and controlling for market dynamics can significantly enhance their predictive power and investment performance [55][56].

Market Performance - The Sci-Tech 50 Index increased by 4.63% over the last five trading days, while the ChiNext Index rose by 2.76%. In contrast, the large-cap value index fell by 0.11%, and the large-cap growth index increased by 2.41% [1] - The construction materials and coal sectors performed well, whereas the banking and telecommunications sectors lagged behind [1] Risk Premium Analysis - The risk premium, defined as the difference between the static PE of the CSI All Share Index and the yield of 10-year government bonds, has shown significant historical extremes. As of April 26, 2022, it reached 4.17%, and on October 28, 2022, it was 4.08%. The latest reading on January 19, 2024, was 4.11%, marking the fifth instance since 2016 where it exceeded 4% [1] - As of July 25, 2025, the risk premium indicator stands at 3.35%, with the two-standard-deviation boundary at 4.76% [1] Valuation Levels - As of July 25, 2025, the CSI All Share Index's PE TTM percentile is at 67%. The Shanghai Stock Exchange 50 and CSI 300 indices are at 68% and 62%, respectively, while the ChiNext Index is at approximately 26%. The CSI 500 and CSI 1000 indices are at 49% and 38%, indicating that the ChiNext Index is relatively undervalued compared to historical averages [2] Long-term Market Trends - The Shenzhen 100 Index has historically experienced bear markets every three years, followed by bull markets. The last adjustment began in Q1 2021, suggesting that the current market has ample time and space for a potential upward cycle [2] Fund Flow and Trading Activity - In the last five trading days, ETF inflows totaled 4.3 billion yuan, and the margin trading balance increased by approximately 36.9 billion yuan. The average daily trading volume across both markets was 181.79 billion yuan [4] AI and Machine Learning Applications - A convolutional neural network (CNN) model has been utilized to analyze graphical price and volume data, mapping learned features to industry themes. The latest focus is on sectors such as non-ferrous metals [3][10]