广发证券资深金工分析师 周飞鹏 SAC: S0260521120003 zhoufeipeng@gf.com.cn 广发证券首席金工分析师 安宁宁 SAC: S0260512020003 anningning@gf.com.cn 广发金工安宁宁陈原文团队 摘要 主要结论： 在前期报告《龙头扩散效应行业轮动之二-优选行业组合构建》中我们构建了月度优选行业轮动组合。而由于部分一级行业暂无 对应ETF作为投资工具，部分时期可能存在无法直接持有特定行业标的的情况。为最大限度获取轮动收益，本报告尝试从选股的角度探讨以 持股来复制或增厚轮动策略收益的方式。 行业个股双驱优选组合月度调仓，2013年以来年化收益33.6%，相对中证500指数年化超额收益 28.3%，IR2.07，相对最大回撤27.8%。 板块行情的底层驱动机制： 受活跃资金挖掘概念主题手法的启发，我们思考板块行情的启动和发展过程或许在微观层面上源自板块内个股上涨的 蔓延与扩散，从最初的行情龙头到概念相关的更多个股，正是行情覆盖范围的逐渐延伸催生了一轮行业上行趋势。我们将此过程称为"龙头扩散效 应"。 收益复制角度： 行业全复制组合收益复制效果最理想，但由于 ...

广发证券资深金工分析师 李豪 lhao@gf.com.cn 广发金工安宁宁陈原文团队 摘要 基于平均真实波幅的ETF网格交易策略： 广发证券首席金工分析师 安宁宁 anningning@gf.com.cn ETF网格交易策略组合： 基于加入择时信号之后的网格交易策略，我们以日度为频率筛选规模、流动性较为可观的ETF构建备选 池，并同样以日频为频率构建ETF网格交易策略组合。长期来看，ETF网格交易策略组合的走势呈现出 较为稳健的绝对收益特征。进一步地，我们设定不同备选池更新周期、ETF换仓周期以及权益配置比例 上限，测试了不同情况下ETF网格交易策略组合的表现。此外，我们基于相同方法，采用常见宽基指数 的对应ETF构建宽基指数ETF网格交易策略组合。历史上来看，该组合的走势同样呈现出较为稳健的绝 对收益特征。 本文中，我们尝试针对ETF构建网格交易策略组合。网格交易策略基于高抛低吸的核心思想，将价格波 动区间划分为多个网格，在价格下跌时买入，在价格上涨时卖出，从价格的波动之中获取利润。具体流 程上，我们从指数的平均真实波幅(ATR)出发，设定具体的网格策略参数；而后，我们基于ATR指标构 建单指数网格交易策略 ...

广发证券资深金工分析师 张钰东 广发证券首席金工分析师 安宁宁 SAC: S0260512020003 anningning@gf.com.cn SAC: S0260522070006 zhangyudong@gf.com.cn 广发金工安宁宁陈原文团队 摘要 最近5个交易日，科创50指数跌3.85%，创业板指跌3.01%，大盘价值涨1.44%，大盘成长跌1.64%，上证50涨0.003%，国证2000代表的小盘跌0.53%，综 合、纺织服饰表现靠前，通信、电子表现靠后。 风险溢价，中证全指静态PE的倒数EP减去十年期国债收益率，权益与债券资产隐含收益率对比，截至2025/11/14指标2.78%，两倍标准差边界为4.74%。 估值水平，截至2025/11/14，中证全指PETTM分位数81%，上证50与沪深300分别为77%、73%，创业板指接近50%，中证500与中证1000分别为62%、 61%，创业板指风格估值相对历史总体处于中位数水平。 使用卷积神经网络对图表化的价量数据与未来价格进行建模，并将学习的特征映射到行业主题板块中。最新配置主题为能源、高股息等，具体包括中证能 源指数、中证高股息策略指数和 ...

广发证券首席金工分析师 安宁宁 SAC: S0260512020003 anningning@gf.com.cn 广发证券资深金工分析师 陈原文 SAC: S0260517080003 chenyuanwen@gf.com.cn 广发证券资深金工分析师 王小康 SAC: S0260525020002 wangxiaokang@gf.com.cn 广发金工安宁宁陈原文团队 摘要 基于神经网络模型构建的择时策略表现优异 ：我们使用前期发现在截面选股上效果优异的AGRU模 型，使用一种新的训练方式，使其可以用于构建择时策略。若使用中证全指作为持仓标的，假设每日开 盘价成交，手续费单边万五（双边千一）的情况，策略年化收益率为15.86%，夏普比率为1.18，最大回 撤为-12.66%。 我们对比了不同的调仓周期下策略表现，发现周度和月度调仓会对策略产生较大负面影响。一方面择时 信号的IC随着周期拉长逐渐下降，一方面预测信号量不足天然降低了容错空间，会对策略稳定性产生影 响。 日度调仓截面ETF轮动策略： 对于截面ETF轮动，我们使用前期报告中的研究成果，调整为每日预测并 调仓后，假设持仓10只ETF，相较于所有ETF ...

摘要 盈利、成长是选股的重要变量： 在2020年8月26日广发金融工程团队发布了关于长线选股系列报告 (二)：《如何挖掘景气向上，持续增长企业》，围绕盈利和成长两大维度，辅助个股质量、价值等维度 对个股进行精选。本篇报告对该策略的表现进行跟踪，以下如无特别说明，本篇报告数据来源均为 Wind。 实证分析： 在回测期内，等权重策略在历史回测期内相对中证800指数年化波动率为13.63%，年化信息 比为1.19。组合从个股持股数量上看，平均每期持股数量约为55只，每期组合内个股的平均流通市值为 140亿元左右；从历史组合每期入选个股的行业分布中可以看出，医药生物、化工、电子、机械设备、 食品饮料等行业个股入选次数较多。 风险提示： 本专题报告所述模型用量化方法通过历史数据统计、建模和测算完成，所得结论与规律在 市场政策、环境变化时可能存在失效风险；策略在市场结构改变时有可能存在策略失效风险。策略在交 易行为改变时存在可能失效风险。 一、引言 在2020年8月26日广发金融工程团队发布了关于长线选股系列报告(二)：《如何挖掘景气向上，持续增 长企业》，在这篇专题报告中，围绕盈利和成长两大维度，辅助个股质量、价值等维 ...

Core Viewpoint - The article emphasizes the growing recognition of index-based investment strategies among investors, highlighting the potential investment opportunities arising from significant changes in index constituents due to periodic rebalancing of major indices like the SSE 50, CSI 300, and CSI 500 [1][4][5]. Group 1: Index Product Scale Statistics - As of October 31, there are 2,294 passive index funds (ETFs and off-exchange passive index funds) with a total scale of 4.5 trillion yuan, and 437 enhanced index funds with a total scale of 265.3 billion yuan, surpassing the scale of equity mixed funds (2.53 trillion yuan) [2][15]. - The leading indices in terms of product tracking scale include the CSI 300, CSI A500, and CSI 500 [19]. Group 2: Historical Adjustment Effects of Index Constituents - From 2019 to mid-2025, stocks added to the index generally outperformed the index in the two weeks prior to their inclusion, while stocks removed from the index underperformed [22][23]. - The average excess return for stocks added to the index in the two weeks before inclusion is 4.89%, with a success rate of 66.67% [24]. Group 3: Latest Adjustment Impact Estimates - For the expected adjustments in December 2025, the SSE 50 is projected to adjust 4 stocks with an estimated passive buy amount of 5.5 billion yuan, the CSI 300 is expected to adjust 10 stocks with an estimated net buy of 24.5 billion yuan, and the CSI 500 is expected to adjust 50 stocks with an estimated buy of 3.3 billion yuan [30][32].

Market Performance - The recent five trading days saw the Sci-Tech 50 Index increase by 0.01%, the ChiNext Index by 0.65%, the large-cap value index by 2.33%, the large-cap growth index by 0.28%, the SSE 50 by 0.89%, and the small-cap index represented by the CSI 2000 by 0.52% [1] - Sectors such as electric equipment and coal performed well, while computer and beauty care sectors lagged behind [1] Valuation Levels - As of November 7, 2025, the static PE of the CSI All Index is at an 82nd percentile, with the SSE 50 and CSI 300 at 77% and 74% respectively, while the ChiNext Index is close to 53% [1] - The valuation of the ChiNext Index is relatively at the historical median level [1] Risk Premium - The risk premium, calculated as the inverse of the static PE of the CSI All Index minus the yield of ten-year government bonds, stands at 2.78% as of November 7, 2025, with a two-standard deviation boundary at 4.74% [1] ETF Fund Flows - In the last five trading days, ETF inflows amounted to 37.2 billion yuan, while margin trading decreased by approximately 700 million yuan [2] Industry Themes - The latest thematic allocation includes banking, energy, and dividends, specifically focusing on indices such as the CSI Bank Index, CSI Energy Index, and CSI Central Enterprises Dividend Index [2][3] Long-term Market Sentiment - The proportion of stocks above the 200-day moving average is being tracked to gauge long-term market sentiment [13] Financing Balance - The financing balance is being monitored to assess market liquidity and investor sentiment [16] Individual Stock Performance - Statistics on individual stock performance year-to-date based on return ranges are being compiled to identify trends [18] Oversold Indices - Observations are being made regarding indices that are considered oversold, indicating potential investment opportunities [20]

Core Viewpoint - The article emphasizes the importance of factor timing in investment strategies, highlighting the need to dynamically select effective factors based on changing market conditions to enhance the stability of multi-factor strategy returns [1][9]. Factor Timing Signals Effectiveness - A total of 92 timing signals were tested, showing an average correlation coefficient of over 15% with the next period's long returns across 77 Alpha factors and 10 Barra style factors. Specifically, deep learning, Level-2, minute frequency, and Barra factors had average correlation coefficients of 17%, 14%, 15%, and 14% respectively, indicating strong predictive power [2][19]. - The deep learning factors such as agru_dailyquote, DL_1, and fimage exhibited average correlation coefficients of 17%, 15%, and 18% respectively, with significant correlations observed in momentum, volatility, liquidity, and market capitalization characteristics [19]. Multi-Signal - Single Factor Timing - To avoid multicollinearity issues, the article employed Partial Least Squares (PLS) for signal aggregation and prediction. The AI image factor fimage achieved a timing success rate of 79%, with an excess annualized return of 8.9% and a Sharpe ratio improvement of 0.67 [2][39]. Multi-Signal - Multi-Factor Timing - The article presented a multi-factor timing strategy that resulted in an annualized return of 37.0% and a Sharpe ratio of 1.72, compared to a non-timed equal-weighted portfolio's annualized return of 20.8% and Sharpe ratio of 0.78. This led to an excess annualized return of 11.6% and a Sharpe ratio improvement of 0.94 [4][5]. Dynamic Multi-Factor Composite - Factor timing can be dynamically integrated into multi-factor composites for strategies like index enhancement. The timing factors in the index enhancement strategies for various indices, including CSI 300 and ChiNext, showed excess annualized returns of 4.56%, 5.98%, 1.08%, 5.67%, and 0.17% compared to the benchmark [5]. Factor Performance Statistics - The article analyzed the performance of 77 Alpha factors and 10 Barra style factors, providing detailed statistics on their returns and predictive capabilities. The results indicated that the factors maintained a strong predictive ability over various time frames [10][19]. Timing Signal Construction - The constructed timing signals fall into four main categories: Momentum, Volatility, Reversal, and Characteristics Spread. Each category has specific methodologies for calculating the signals, focusing on historical returns, volatility, and other characteristics [11][12][13][15][17][18].

Core Viewpoint - The article emphasizes the growing recognition of index-based investment among investors, highlighting the potential investment opportunities arising from significant changes in index constituents due to the periodic rebalancing of major indices like the SSE 50, CSI 300, and CSI 500 [1][4]. Group 1: Index Fund Growth - The total scale of passive index funds (including ETFs and off-market passive index funds) reached 4.5 trillion yuan as of October 31, with 2,294 funds, while enhanced index funds totaled 265.3 billion yuan, surpassing the scale of equity mixed funds at 2.53 trillion yuan [2][15]. - The total scale of equity ETFs grew from approximately 200 billion yuan in 2014 to 3.72 trillion yuan by October 2025, indicating significant growth [15]. Group 2: Historical Adjustment Effects of Index Constituents - Historical analysis from 2019 to mid-2025 shows that stocks added to indices tend to outperform the index in the two weeks prior to their inclusion, while those removed tend to underperform [2][24]. - The average excess return for stocks added to the index in the two weeks before inclusion was 4.89%, with a success rate of 66.67% [25]. Group 3: Latest Adjustment Impact Estimation - The expected adjustments for December 2025 indicate that the SSE 50 will adjust 4 stocks with an estimated passive buy amount of 5.5 billion yuan, the CSI 300 will adjust 10 stocks with an estimated net buy of 24.5 billion yuan, and the CSI 500 will adjust 50 stocks with an estimated buy of 3.3 billion yuan [3][33].

Macro and Technical Perspectives on Asset Allocation - The macro perspective indicates a bearish outlook on equity assets, while the technical perspective shows an upward trend with moderate valuation and a state of capital outflow [1][5][20] - For bonds, the macro perspective is bullish, but the technical perspective indicates a downward trend [1][5][20] - Industrial products are viewed negatively from a macro perspective, with a downward price trend also noted technically [1][5][20] - Gold assets are favored in the macro view, with a technical upward price trend [1][5][20] Asset Performance Tracking - The fixed ratio + macro indicators + technical indicators combination yielded a return of 10.51% in 2025, with an annualized return of 12.05% since April 2006 [2][21] - The volatility-controlled + macro indicators + technical indicators combination achieved a return of 15.69%, while the risk parity + macro indicators + technical indicators combination returned 6.99% [2][30] Asset Class Analysis - Equity assets are currently under pressure from macro indicators, while technical indicators suggest an upward trend but with capital outflow [20][21] - Bond assets are supported by macro indicators, but technical indicators show a downward trend [20][21] - Industrial products face macro headwinds and technical downward trends [20][21] - Gold assets benefit from favorable macro indicators and an upward technical trend [20][21] Valuation and Capital Flow Indicators - The equity risk premium (ERP) for the CSI 800 index is at 53.94%, indicating a moderate valuation level [13][14] - The latest capital flow indicator for equity assets shows a net outflow of 316 billion, reflecting a capital outflow state [16][17] Summary of Asset Class Scores - The overall scores for asset classes based on macro and technical indicators show equities at 1, bonds at 3, industrial products at -2, and gold at 2 [19][20] - The combined analysis suggests a bearish outlook for equities and industrial products, while bonds and gold are viewed positively [20][21]