Report Summary 1. Report Industry Investment Rating - Not provided in the given content 2. Core Viewpoints of the Report - The strategy is based on the research reports "Strategy Portfolio Report under Industry Rotation: Quantitative Analysis from the Perspective of Industry Style Continuity and Switching" (20241007) and "Research on the Overview and Allocation Methods of the Stock - type ETF Market: Taking the ETF Portfolio Based on the Industry Rotation Strategy as an Example" (20241013) to construct a strategy portfolio of industry and theme ETFs [2] - In the week of 20250811, the model recommends allocating sectors such as joint - stock banks, games, and semiconductors. In the next week, the strategy will newly hold products like Game ETF, Science and Technology Innovation Chip Design ETF, and Satellite ETF, and continue to hold products like Bank ETF, Financial Real Estate ETF, and Gold Stock ETF [2] - As of last weekend, some ETFs and the trading timing signals of the underlying indexes gave daily or weekly risk warnings [2] 3. Summary by Relevant Catalogs Performance Tracking - During the period from 20250804 to 20250808, the cumulative net return of the strategy was about 2.62%, and the excess return relative to the CSI 300 ETF was about 1.41% [3] - From October 14, 2024, to the present, the cumulative out - of - sample return of the strategy was about 7.08%, and the cumulative excess relative to the CSI 300 ETF was about - 0.79% [3] Future 1 - Week Recommended ETFs (20250811 - 20250815) | Fund Code | ETF Name | Holding Status | ETF Market Value (billion yuan) | Heavy - Positioned Shenwan II Industry and Weight | Weekly Timing Signal | Daily Timing Signal | | --- | --- | --- | --- | --- | --- | --- | | 512800 | Bank ETF | Continue to hold | 151.38 | Joint - stock banks (44.73%) | 1 | - 1 | | 159869 | Game ETF | Transfer in | 73.17 | Games (81.29%) | 1 | 1 | | 588780 | Science and Technology Innovation Chip Design ETF | Transfer in | 2.77 | Semiconductors (95.73%) | 1 | 1 | | 159940 | Financial Real Estate ETF | Continue to hold | 7.99 | Securities (29.12%) | 1 | - 1 | | 517520 | Gold Stock ETF | Continue to hold | 46.34 | Precious metals (41.51%) | 1 | 1 | | 510000 | Central Enterprise ETF | Continue to hold | 1.21 | State - owned large - scale banks (18.11%) | 1 | 1 | | 512690 | Wine ETF | Continue to hold | 152.39 | Baijiu (85.37%) | - 1 | - 1 | | 159206 | ZETF | Transfer in | 1.33 | Military electronics II (34.22%) | 1 | 1 | | 159786 | VRETF | Transfer in | 1.32 | Optoelectronics (26.64%) | 1 | 1 | | 159652 | Non - ferrous 50 ETF | Transfer in | 5.21 | Industrial metals (49.34%) | 1 | 1 | [9] Near 1 - Week ETF Holdings and Performance (20250804 - 20250808) | Fund Code | Current Holding Status | ETF Name | ETF Market Value (billion yuan) | Near 1 - Week Increase/Decrease (%) | | --- | --- | --- | --- | --- | | 562550 | - | Green Power ETF | 1.21 | 1.50 | | 512800 | Continue to hold | Bank ETF | 151.38 | 1.99 | | 512690 | Continue to hold | Wine ETF | 152.39 | 1.06 | | 159768 | - | Real Estate ETF | 6.13 | 2.14 | | 159940 | Continue to hold | Financial Real Estate ETF | 7.99 | 1.41 | | 515220 | Transfer out | Coal ETF | 80.20 | 3.78 | | 159996 | Transfer out | Home Appliance ETF | 12.72 | 2.55 | | 510060 | Continue to hold | Central Enterprise ETF | 1.21 | 1.42 | | 516550 | Transfer out | Agricultural ETF | 1.87 | 1.76 | | 517520 | Continue to hold | Gold Stock ETF | 46.34 | 8.91 | | - | ETF Portfolio Average Return | - | - | 2.62 | | 510300 | - | CSI 300 ETF | 3819.72 | 1.21 | | - | ETF Portfolio Excess Return | - | - | 1.41 | [10]

Group 1 - The report identifies six main categories of public quantitative products, totaling 919 products with a combined scale of 386.87 billion yuan, with the largest categories being index enhancement, active quantitative, and absolute return funds [4][10][11] - Active quantitative funds have shown a higher correlation with quality, growth, and small-cap factors in 2020, shifting to a stronger correlation with small-cap factors after 2022 [4][22] - The report outlines four main strategies for active quantitative funds, including equity fund enhancement, SmartBeta style strategies, industry rotation/all-industry quantitative stock selection strategies, and quantitative strategies from active equity teams [4][12][17] Group 2 - The report details the distribution of public quantitative products by strategy, highlighting that index enhancement funds account for 374 products with a scale of 194.32 billion yuan, while active quantitative funds consist of 343 products totaling 91.39 billion yuan [10][11] - The active quantitative fund strategies include equity fund enhancement represented by products like Baodao Qihang and Baodao Yuhang, SmartBeta strategies focusing on small-cap and dividend strategies, and industry rotation products like Huashan Event-Driven Quantitative Strategy [4][12][17] - The report emphasizes the differences in active quantitative product layouts among various fund companies, with specific companies focusing on different strategies such as full-industry quantitative stock selection or style funds [13][14] Group 3 - The report lists the top 20 active quantitative products by scale, with notable products including Zhao Shang Quantitative Selection A and Guojin Quantitative Multi-Factor A, highlighting their respective strategies and performance metrics [15][24] - It notes that many of the top-performing active quantitative products in 2025 have a tendency to invest in small-cap stocks, indicating a market trend towards smaller market capitalization [25][26] - The report also discusses the environmental adaptability of different strategies within active quantitative funds, indicating varying performance based on market conditions [26]

Group 1 - The report suggests that the current market adjustment is not expected to be sustained, presenting a potential short-term accumulation window, driven by high-growth sectors such as PCB, optical modules, and innovative pharmaceuticals [2][3] - It is recommended to moderately increase positions in leading sectors in August, while also paying attention to potential rotation directions [3] - The report emphasizes the importance of semiconductor and medical device industries, which are currently at relatively low levels and closely related to the main logic of the current market trend, offering dual advantages of valuation recovery and sustained growth [6][8] Group 2 - The report highlights the automotive industry as worthy of attention due to its low valuation and potential for marginal improvement under the influence of policies aimed at reducing internal competition [6][8] - The report identifies several low-position industries with marginal improvements, including electric motors, automotive, and environmental protection, indicating positive trends in key indicators such as export amounts and production levels [7][8] - The report notes that the political bureau meeting's content did not meet expectations, leading to market corrections in sectors like anti-involution, real estate, and cyclical industries [6][8]

Core Insights - The report highlights the current industry allocation positions of the Bank of China multi-strategy system, with the highest allocations in Computer (9.7%), Power Equipment and New Energy (8.0%), and Non-ferrous Metals (7.9) [1] - The average weekly return for the CITIC primary industries is 0.8%, with the best-performing sectors being Communication (7.1%), Pharmaceuticals (4.3%), and Computers (4.0%) [3][11] - The report indicates that the composite strategy has achieved a cumulative return of 12.6% year-to-date, outperforming the CITIC primary industry equal-weight benchmark by 1.9% [3] Industry Performance Review - The best-performing sectors this week are Communication (7.1%), Pharmaceuticals (4.3%), and Computers (4.0%), while the worst-performing sectors are Banks (-2.8%), Real Estate (-2.3%), and Coal (-2.2%) [3][11] - The average monthly return over the past month for the CITIC primary industries is 7.0% [3][11] Valuation Risk Warning - The report employs a valuation warning system based on the PB ratio over the past six years, identifying sectors with high valuation risks. Currently, the sectors of Retail, Automotive, and Media are flagged for high valuation warnings as their PB ratios exceed the 95th percentile [13][14] Single Strategy Performance - The top three industries based on the high prosperity industry rotation strategy (S1) are Computer, Non-ferrous Metals, and Power Equipment and New Energy [16][17] - The implied sentiment momentum tracking strategy (S2) ranks the top three industries as Computer, Communication, and Machinery [20][22] - The macro style rotation strategy (S3) identifies the top six industries as Comprehensive Finance, Computer, Media, National Defense Industry, Electronics, and Comprehensive [24][25] Strategy Adjustments - The report notes that only two weekly strategies made adjustments this week, with an overall slight change in positions. The composite strategy increased exposure to midstream non-cyclical sectors while reducing exposure to financial sectors [3]

Core Insights - The report emphasizes the integration of AI in asset allocation, enhancing the predictive capabilities of stock and bond performance through a dynamic weighting system [2][3] - The AI-driven model has successfully predicted market trends, including the recent performance of value stocks outperforming growth stocks in March and April [3] - Predictions for 2025 indicate that bond assets will maintain relative advantages, while stock market performance is expected to stabilize at the bottom in Q3 and slightly recover in Q4 [3] Asset Allocation Framework - The AI-enabled research system combines five major cycles to predict stock and bond performance, with a current high weighting of 55% on the monetary credit framework [2][3] - The allocation for domestic assets in July shows: 12.64% in equities, 3.58% in dividends, 76.45% in bonds, and 7.33% in gold, with adjustments compared to traditional risk parity models [4] - For overseas markets, the allocation includes: France 15.62%, Germany 14.85%, the US 20.24%, Japan 16.44%, Hong Kong 11.50%, and India 22.35%, with slight adjustments in France, Germany, and Hong Kong [4] Industry Rotation Strategy - The AI-driven industry rotation strategy has significantly improved performance metrics, achieving a 420% increase in the Sharpe ratio and a 41% reduction in maximum drawdown compared to traditional strategies [5] - The latest industry outlook for Q3 suggests overweight positions in machinery, comprehensive sectors, and electronics, while maintaining standard positions in automotive, communication, and construction, and underweighting banking and retail [5]

Core Insights - The report highlights a focus on specific stocks for June, including 顺丰控股 (SF Holding), 安集科技 (Anji Technology), and 佰仁医疗 (Bairen Medical) among others, indicating potential investment opportunities in these companies [1] - The overall market indices showed slight declines, with the Shanghai Composite Index closing at 3448.45, down 0.22% [1] Strategy Research - The report discusses a traditional multi-factor scoring industry rotation strategy that prioritizes low valuation, low crowding, and upward economic momentum, achieving an annualized return of 19.64% during the backtest period from April 1, 2014, to June 6, 2025, compared to a benchmark return of 7.55% [2][8] - The strategy involves selecting two single factors from four dimensions: valuation, quality, liquidity, and momentum, and forming a composite factor through equal weighting [8][9] Mechanical Equipment Sector - 芯碁微装 (Chipbond Technology) announced a new contract worth 146 million yuan, representing approximately 15% of its projected 2024 revenue, indicating strong demand driven by the AI infrastructure boom [10][11] - The company reported a revenue of 242 million yuan in Q1 2025, with a quarter-over-quarter increase of 3% and a year-over-year increase of 22%, alongside a gross margin of 41.3% [11][12] - The AI infrastructure trend is expected to significantly boost the demand for high-end PCB products, with major tech companies like Meta and Microsoft increasing their capital expenditures for AI-related infrastructure [12]

Quantitative Models and Construction Methods Single Strategy Models - **Model Name**: High Prosperity Industry Rotation Strategy **Construction Idea**: Tracks industry profitability expectations using multi-factor models based on analysts' consensus data to select industries with upward profitability trends [13][15][16] **Construction Process**: 1. Constructs three types of factors: - Type 1: Long-term profitability factors (e.g., ROE_FY2, ROE_FY1) - Type 2: Quarterly changes in profitability (e.g., EPS_F2_qoq, EPS_F3_mom) - Type 3: Monthly changes in profitability (e.g., EPS_F3_qoq_d1m) 2. Filters industries with extreme valuations using PB percentile thresholds [30] 3. Selects top 3 industries based on composite factor rankings and allocates equally [21][30] **Evaluation**: Demonstrates strong performance in tracking industry cycles and avoiding valuation bubbles [13][26] - **Model Name**: Implicit Sentiment Momentum Strategy **Construction Idea**: Captures "unverified sentiment" by removing the relationship between turnover rate changes and returns, aiming to identify market sentiment-driven opportunities [32][33] **Construction Process**: 1. Uses OLS regression to remove "expected sentiment" from daily industry returns, leaving residuals as "unverified sentiment" [34] 2. Constructs momentum factors based on cumulative "unverified sentiment" returns over various time windows (e.g., 1 month, 12 months) [35] 3. Enhances the strategy by neutralizing fundamental impacts, adjusting for volatility, and applying composite factor methods [36] **Evaluation**: Effectively captures sentiment-driven market dynamics ahead of fundamental data releases [32][37] - **Model Name**: Macro Indicator Style Rotation Strategy **Construction Idea**: Uses macroeconomic indicators to predict industry styles (e.g., value, momentum) and maps them to industry selection [43][44] **Construction Process**: 1. Constructs macro indicators (e.g., PMI, CPI, M1) using historical positioning, surprise, and marginal change metrics [48][49] 2. Builds style factors (e.g., Value, Beta, Momentum) based on industry exposures [50][51] 3. Maps style predictions to industry scores and selects top industries [61] **Evaluation**: Addresses limitations of traditional top-down models by incorporating style-based predictions [43][61] - **Model Name**: Mid-to-Long-Term Momentum Reversal Strategy **Construction Idea**: Explores the "momentum-reversal" structure in industry returns, combining short-term momentum and long-term reversal factors [70][71] **Construction Process**: 1. Constructs momentum factors based on single-month returns and reversal factors based on multi-month returns (e.g., 12-month momentum, 24-36 month reversal) [76][78] 2. Combines factors using rank-weighted methods and adjusts for turnover rates [80][85] **Evaluation**: Balances short-term trends and long-term recovery opportunities effectively [70][84] - **Model Name**: Fund Flow Industry Rotation Strategy **Construction Idea**: Tracks institutional and tail-end fund flows to identify industry momentum [91][92] **Construction Process**: 1. Constructs "institutional trend strength factors" based on net buy amounts [93][94] 2. Constructs "tail-end inflow strength factors" based on post-14:30 net inflow data [96][103] 3. Combines factors and excludes high-concentration industries [100][101] **Evaluation**: Enhances stability by avoiding crowded trades [91][101] - **Model Name**: Financial Report Failure Reversal Strategy **Construction Idea**: Utilizes mean-reversion characteristics of long-term effective financial factors after short-term failures [108][109] **Construction Process**: 1. Constructs financial factors (e.g., ROA, YOY) using profit and balance sheet data [110][114] 2. Identifies "long-term effective factors" and "recently failed factors" based on rolling windows [116][117] 3. Combines factors using zscore methods [117] **Evaluation**: Captures recovery opportunities in temporarily underperforming factors [108][118] - **Model Name**: Traditional Low-Frequency Multi-Factor Scoring Strategy **Construction Idea**: Combines factors from four dimensions (momentum, valuation, liquidity, quality) for quarterly industry rotation [122][123] **Construction Process**: 1. Selects top-performing factors from each dimension (e.g., 1-year momentum, ROE_TTM) [124][125] 2. Combines factors using rank-weighted methods [135] 3. Filters industries with low weights in the CSI 800 index [135] **Evaluation**: Suitable for long-term holding with robust risk control [122][129] Composite Strategy Models - **Model Name**: Volatility-Controlled Composite Strategy **Construction Idea**: Allocates funds across single strategies based on inverse negative volatility [138][139] **Construction Process**: 1. Calculates negative volatility for each strategy over a rolling window (e.g., 63 days) [139][140] 2. Allocates funds proportionally to inverse negative volatility [139][147] 3. Adjusts allocation frequencies to match individual strategy cycles (weekly, monthly, quarterly) [141][146] **Evaluation**: Balances risk and return effectively, achieving high annualized excess returns [138][144] --- Model Backtest Results Single Strategy Results - **High Prosperity Strategy**: Annualized excess return 16.69%, max drawdown -12.95%, IR 1.29 [26] - **Implicit Sentiment Strategy**: Annualized excess return 18.61%, max drawdown -17.83%, IR 1.04 [37] - **Macro Style Strategy**: Annualized excess return 7.01%, max drawdown -23.46%, IR 0.30 [63] - **Momentum Reversal Strategy**: Annualized excess return 11.42%, max drawdown -14.91%, IR 0.77 [84] - **Fund Flow Strategy**: Annualized excess return 11.64%, max drawdown -12.16%, IR 0.96 [101] - **Financial Report Strategy**: Annualized excess return 9.13%, max drawdown -10.54%, IR 0.87 [118] - **Low-Frequency Multi-Factor Strategy**: Annualized excess return 12.00%, max drawdown -13.25%, IR 0.91 [129] Composite Strategy Results - **Volatility-Controlled Composite Strategy**: Annualized excess return 12.2%, max drawdown -6.8%, IR 1.80 [144][147]

Group 1 - The core concept of industry rotation strategy is to profit from structural market trends by switching between different industry sectors to maximize investment returns or mitigate systemic risks [1] - Industry rotation strategies are particularly effective in market environments where there are significant differences in returns among various industries during the same time period [1] Group 2 - Utilizing ETFs for industry rotation offers several advantages, including simplicity of operation, as investors can easily gain exposure to a basket of stocks representing an industry without the need to buy multiple individual stocks [2] - ETFs provide transparency in holdings, with daily disclosures of their portfolios, allowing investors to clearly understand their investments, unlike actively managed equity funds which may have delayed reporting [2] - The cost-effectiveness of ETFs is highlighted, as they can be traded at any time during the trading day, and generally have lower fees compared to traditional actively managed funds, enhancing investor returns [2] Group 3 - The method for constructing an industry rotation strategy using ETFs involves scoring ETFs based on a series of industry selection indicators, selecting those with higher scores for allocation, and making regular adjustments [3] - The scoring process includes evaluating the performance of the underlying stocks in terms of fundamentals, technicals, and capital flows, leading to an overall score for the ETF [3]

当ETF赛道深陷费率战与规模焦虑时，中证A500指数却以另类姿态撕开市场——这只诞生即被贴上"新锐"标 签的宽基指数，凭借对科创属性与中小市值的倾斜性覆盖，成为近两年机构博弈"贝塔收益"的主战场。 除了密集成立的指数基金以外，截至今年4月，全市场已有26只指数增强产品参与竞逐，不同产品之间分化 剧烈：两只成立时间间隔不到一个月的A500指数增强基金，目前的超额收益差值已经接近10%。 归根结底，A500指数"市值+行业双轮筛选"的编制原则，使得成份股市值和流动性分层显著，为量化模型留 足了"翻石头"的空间。因此，在选择A500指数增强基金时，基金经理的投资能力与增强策略变得至关重 要。 华安基金量化投资部助理总监、基金经理张序的突围密码，藏在八年磨一剑的"行业轮动+多因子"双擎模型 里。通过对行业轮动的深度理解和持续迭代，其管理的华安事件驱动量化基金自2020年执掌以来，连续五 年跑赢偏股混基指数，年化超额收益达9.3%，无论在公募量化还是主动股基均排名前1%。 而当市场还在争论主动量化与被动投资的边界时，华安基金已悄然完成中证A500产品线的战术合围。继 2024年精准卡位A500ETF之后，再次推出了由张 ...

广发证券资深金工分析师 李豪 lhao@gf.com.cn 广发证券首席金工分析师 安宁宁 anningning@gf.com.cn 广发金工安宁宁陈原文团队 摘要 大语言模型在金融领域的应用： 近年来，人工智能技术的快速发展推动了大语言模型(LLMs)的革新。作为最前沿的技术之一，大语言 模型正在广泛应用于各行各业。金融行业作为一个高度依赖数据分析和信息处理的领域，对先进的人工 智能技术有着极大的需求。而LLMs凭借其强大的文本理解能力、信息提取能力以及推理和预测能力， 正在逐步改变传统的金融分析和决策方式，为投资管理、市场分析、风险控制等多个领域带来了新的机 遇。 DeepSeek定量解析基金季报行业观点及行业轮动策略构建： 本文中，我们尝试通过DeepSeekV3模型，对于基金季报观点文本中的行业观点进行定量解析，并以此 出发构建行业轮动策略。具体来看，首先我们筛选存续时间较长的主动型权益基金样本，并提取样本基 金不同季度报告期季报中的观点部分文本；而后我们将观点文本输入至DeepSeek模型，加入特定提示 词控制输出的格式，并基于输出结果构建基金季报行业观点指标；最后我们基于基金季报行业观点指标 及观 ...