报告导读： 本文主要考察根据因子历史长短期收益、波动率等数据，采用 PLS 模型预期 因子收益在因子加权中的应用，包括单因子多策略、以及多因子单策略两个维度。 在 20 个单因子 top100 组合配置中，对波动率最大的 5 个因子组合采用 PLS 模型预期收益确定权重，相比于收益均值加权方式，年化收益提升约 4.0% （ 2018.01-2025.11 ，下同）；相较于等权方式，年化收益提升 6.6% 。 风格组合配置维度，我们构建了 6 个基础组合： 1 个红利优选、 1 个成长期优选、两个小市值组合、以及两个风格相对均衡的组合 --PB 盈利和 GARP 组 合。对波动较大组合采用 PLS 模型预期超额收益加权，相比于超额收益均值加权方式，年化收益提升 3.3% ；相较于等权方式，年化收益提升 3.9% 。 量化固收 + 策略中的应用，股票端： PLS 预期收益加权复合风格组合；债券端：中证短债指数；构建 10% 股票仓位的固收 + 量化策略， 2018 年以来年 化收益 5.6% ，年化波动率 2.6% ，信息比 2.17 ；权益比例为 20% 时，策略年化收益 8.4% ，年化波动率 5.1% ，信 ...

- The report tracks the performance of single factors and major factor categories in quantitative stock selection models across different stock pools (CSI 300, CSI 500, CSI 1000, CSI 2000, and CSI All Share). It highlights the excess returns of factors over different time periods, such as weekly, monthly, and yearly[7][28][30] - Single factors with strong weekly excess returns in the CSI 300 stock pool include "120-day change in analysts' forecasted net profit FY3" (2.11%), "EPS 120-day change FY3" (2.05%), and "90-day institutional earnings forecast adjustment" (1.96%)[30] - In the CSI 500 stock pool, factors with strong weekly excess returns include "standardized unexpected single-quarter ROA with drift" (1.18%), "standardized unexpected single-quarter net profit with drift" (1.15%), and "standardized unexpected P/E ratio (parent company) with drift" (1.15%)[31] - For the CSI 1000 stock pool, factors with strong weekly excess returns include "single-quarter non-recurring ROA change" (1.26%), "1-minute path momentum" (1.15%), and "20-day intraday return" (1.02%)[32] - In the CSI 2000 stock pool, factors with strong weekly excess returns include "60-day shareholding ratio change" (1.74%), "5-day shareholding ratio change" (1.26%), and "EP 60-day change" (1.12%)[33] - Major factor categories with strong weekly excess returns in the CSI 300 stock pool include "analysts' surprise" (1.96%), "growth" (1.5%), and "analysts" (1.24%). For the full year, the best-performing categories are "profitability" (33.2%), "analysts' surprise" (30.25%), and "growth" (29.73%)[37][38][40] - In the CSI 500 stock pool, the best-performing major factor categories for the year are "growth" (16.68%), "analysts" (10.84%), and "analysts' surprise" (8.32%)[42][43] - For the CSI 1000 stock pool, the top-performing major factor categories for the year are "growth" (18.41%), "analysts" (10.99%), and "analysts' surprise" (10.86%)[47][48] - In the CSI 2000 stock pool, the best-performing major factor categories for the year are "market capitalization" (22.73%), "analysts' surprise" (20.54%), and "growth" (20.25%)[52][55] - The CSI All Share stock pool shows the best-performing major factor categories for the year as "market capitalization" (43.82%), "growth" (26.45%), and "analysts' surprise" (23.55%)[57][58] - The report also tracks the performance of index enhancement strategies based on multi-factor stock selection models. For the CSI 300 stock pool, the strategy achieved a weekly return of 2.38% and an annual return of 27.19%, with an excess return of 8.84% and a maximum drawdown of -3.15%[59][60] - For the CSI 500 stock pool, the strategy achieved a weekly return of 3.59% and an annual return of 31.54%, with an excess return of 1.28% and a maximum drawdown of -4.76%[60] - The CSI 1000 stock pool's strategy achieved a weekly return of 3% and an annual return of 42.2%, with an excess return of 14.54% and a maximum drawdown of -5.59%[66] - The CSI 2000 stock pool's strategy achieved a weekly return of 1.76% and an annual return of 63.87%, with an excess return of 27.31% and a maximum drawdown of -5.23%[66]

Quantitative Models and Construction Model Name: DTLC_RL (Decoupled Temporal Contrastive Learning with Reinforcement Learning) - **Model Construction Idea**: The model aims to combine the nonlinear predictive power of deep learning with interpretability by decoupling feature spaces, enhancing representation through contrastive learning, ensuring independence via orthogonal constraints, and dynamically fusing spaces using reinforcement learning[2][11][12] - **Model Construction Process**: - **Feature Space Decoupling**: Three orthogonal latent spaces are constructed to capture market systemic risk (β space), stock-specific signals (α space), and fundamental information (θ space). Each space is equipped with a specialized encoder: TCN for β space, Transformer for α space, and gated residual MLP for θ space[11][12][92] - **Contrastive Learning**: Introduced within each space to enhance robustness by constructing positive and negative sample pairs based on return similarity. The InfoNCE loss function is used to maximize the similarity of positive pairs while minimizing that of negative pairs: $$L_{\mathrm{InfotNCE}}=-E\left[l o g~\frac{e x p\left(f(x)^{\top}f(x^{+})/\tau\right)}{e x p\left(f(x)^{\top}f(x^{+})/\tau\right)+\sum_{i=1}^{N-1}~e x p\left(f(x)^{\top}f(x_{i}^{-})/\tau\right)}\right]$$ where \(f(x)\) is the feature representation, \(x^+\) is the positive sample, \(x^-\) is the negative sample, and \(\tau\) is the temperature parameter[55][56] - **Orthogonal Constraints**: A loss function is added to ensure the outputs of the three spaces are statistically independent, reducing multicollinearity and enhancing interpretability[12][104] - **Reinforcement Learning Fusion**: A PPO-based reinforcement learning mechanism dynamically adjusts the weights of the three spaces based on market conditions. The reward function includes components for return correlation, weight stability, and weight diversification: $$r_{t}=R_{t}^{I C}\big(\widehat{y_{t}},y_{y}\big)+\lambda_{s}R_{t}^{s t a b l e}+\lambda_{d}R_{t}^{d i v}$$ The PPO optimization process includes GAE advantage estimation and a clipped policy loss: $$L^{C L P}=E\left[\operatorname*{min}(r\dot{A},c l i p(r,1-\varepsilon,1+\varepsilon)\dot{A})\right]$$[58][120][121] - **Model Evaluation**: The DTLC_RL model demonstrates strong predictive power and interpretability, with dynamic adaptability to market conditions[2][12][122] Model Name: DTLC_Linear - **Model Construction Idea**: A baseline model for comparison, using a linear layer to fuse the three feature spaces[98][100] - **Model Construction Process**: - The encoded information from the three spaces is concatenated and passed through a linear layer with a Softmax activation to generate fusion weights. The model is trained with a multi-task loss function, including IC maximization, contrastive learning loss, and orthogonal constraints[98][104] - **Model Evaluation**: Provides a benchmark for evaluating the contribution of reinforcement learning in DTLC_RL[98][103] Model Name: DTLC_Equal - **Model Construction Idea**: A simpler baseline model that equally weights the three feature spaces without dynamic adjustments[98] - **Model Construction Process**: The outputs of the three spaces are directly averaged to generate predictions[98] - **Model Evaluation**: Serves as a control group to assess the benefits of dynamic weighting in DTLC_RL[98][103] --- Model Backtesting Results DTLC_RL - **IC**: 0.1250[123] - **ICIR**: 4.38[123] - **Top 10% Portfolio Annualized Return**: 34.77%[123] - **Annualized Volatility**: 25.41%[123] - **IR**: 1.37[123] - **Maximum Drawdown**: 40.65%[123] - **Monthly Turnover**: 0.71X[123] DTLC_Linear - **IC**: 0.1239[105] - **ICIR**: 4.25[105] - **Top 10% Portfolio Annualized Return**: 32.95%[105] - **Annualized Volatility**: 24.39%[105] - **IR**: 1.35[105] - **Maximum Drawdown**: 35.94%[105] - **Monthly Turnover**: 0.76X[105] DTLC_Equal - **IC**: 0.1202[105] - **ICIR**: 4.06[105] - **Top 10% Portfolio Annualized Return**: 32.46%[105] - **Annualized Volatility**: 25.29%[105] - **IR**: 1.28[105] - **Maximum Drawdown**: 40.65%[105] - **Monthly Turnover**: 0.71X[105] --- Quantitative Factors and Construction Factor Name: Beta_TCN - **Factor Construction Idea**: Captures market systemic risk by quantifying stock sensitivity to common risk factors like macroeconomic fluctuations and market sentiment[67] - **Factor Construction Process**: - Five market-related features are selected, including beta to market returns, volatility sensitivity, liquidity beta, size exposure, and market sentiment sensitivity[72] - A TCN encoder processes 60-day time-series data, using dilated causal convolutions to capture short- and medium-term trends. The output is a 32-dimensional vector representing systemic risk features[68] - **Factor Evaluation**: Demonstrates moderate stock selection ability and effectively captures market-related information[73] Factor Name: Alpha_Transformer - **Factor Construction Idea**: Extracts stock-specific alpha signals from price-volume time-series data[76] - **Factor Construction Process**: - Thirteen price-volume features are encoded using a multi-scale Transformer model, with separate layers for short-, medium-, and long-term information. Outputs are fused using a gated mechanism and passed through a fully connected layer for return prediction[77][78] - **Factor Evaluation**: Exhibits strong predictive power and stock selection ability, with relatively low correlation to market benchmarks[81][82] Factor Name: Theta-ResMLP - **Factor Construction Idea**: Focuses on fundamental information to assess financial safety margins and risk resistance[88] - **Factor Construction Process**: - Eight core financial indicators, including PE, PB, ROE, and dividend yield, are encoded using a gated residual MLP. The architecture includes input projection, gated residual blocks, and a final output layer[92] - **Factor Evaluation**: Provides stable stock selection performance with lower turnover and drawdown compared to other spaces[95][96] --- Factor Backtesting Results Beta_TCN - **IC**: 0.0969[73] - **ICIR**: 3.73[73] - **Top 10% Portfolio Annualized Return**: 27.73%[73] - **Annualized Volatility**: 27.19%[73] - **IR**: 1.02[73] - **Maximum Drawdown**: 45.80%[73] - **Monthly Turnover**: 0.79X[73] Alpha_Transformer - **IC**: 0.1137[81] - **ICIR**: 4.19[81] - **Top 10% Portfolio Annualized Return**: 32.66%[81] - **Annualized Volatility**: 23.04%[81] - **IR**: 1.42[81] - **Maximum Drawdown**: 27.59%[81] - **Monthly Turnover**: 0.83X[81] Theta-ResMLP - **IC**: 0.0485[95] - **ICIR**: 1.87[95] - **Top 10% Portfolio Annualized Return**: 23.88%[95] - **Annualized Volatility**: 23.96%[95] - **IR**: 0.99[95] - **Maximum Drawdown**: 37.41%[95] - **Monthly Turnover**: 0.41X[95]

- The report discusses multiple quantitative timing models for A-shares, including short-term, medium-term, and long-term models. The short-term models include the Volume Model (neutral for all broad-based indices), Feature Institutional Model (bullish), Feature Volume Model (bearish), and Smart Algorithm Models (bullish for CSI 300, bearish for CSI 500)[1][12][77]. Medium-term models include the Limit-Up-Limit-Down Model (neutral) and the Up-Down Return Difference Model (bullish for all broad-based indices)[13][78]. The long-term model, the Long-Term Momentum Model, is bullish[14][79]. Comprehensive models like the A-Share Comprehensive Weapon V3 Model and the A-Share Comprehensive CSI 2000 Model are bearish[15][80] - For Hong Kong stocks, the medium-term models include the Turnover-to-Volatility Model (bullish) and the Hang Seng Index Up-Down Return Difference Model (neutral)[16][81] - The report emphasizes that timing strategies are not achieved through a single model but require a multi-cycle, multi-strategy system. It highlights the use of price-volume, acceleration, trend, momentum, and limit-up-limit-down perspectives to construct eight major models for market timing[9] - Backtesting results for the Double-Bottom Pattern show a weekly return of 0.29%, outperforming the Shanghai Composite Index by 0.27% during the same period. Since December 31, 2020, the cumulative return of the Double-Bottom Pattern portfolio is 11.79%, slightly underperforming the Shanghai Composite Index's cumulative return of 12.02%[44] - Backtesting results for the Cup-and-Handle Pattern show a weekly return of 0.3%, outperforming the Shanghai Composite Index by 0.27% during the same period. Since December 31, 2020, the cumulative return of the Cup-and-Handle Pattern portfolio is 9.27%, underperforming the Shanghai Composite Index's cumulative return of 12.02%[44]

Report Industry Investment Rating No information about the report industry investment rating is provided in the content. Core Viewpoints of the Report - The issuance speed of US ETFs at the end of the year has increased again, with multiple quantitative enhancement products being issued [2][7]. - The capital inflow of US ETFs has remained above $40 billion, and the risk appetite of capital has remained at a high level [2][13]. - Stock long - short and other alternative strategies of US ETFs have performed well [2][19]. - The redemption pressure of US non - money mutual funds in October 2025 was still high, and domestic stock funds and hybrid products have continued to experience outflows recently, while bond funds have seen a slight inflow [2][20]. Summary by Relevant Catalogs 1. US ETF Innovation Products: Multiple Quantitative Enhancement Products Issued - Last week, 43 new products were issued in the US, including 6 individual stock leverage products and 3 digital currency - related products [2][7]. - Motley Fool issued 3 single - factor ETFs, each holding about 150 stocks [9]. - BlackRock's quantitative team, NEOS, Hedgeye, Global X, Franklin Templeton, Sterling Capital, and Columbia all issued different types of ETFs last week, with many using quantitative strategies [10][11]. 2. US ETF Dynamics 2.1 US ETF Capital: All Types of Assets Maintain Inflows - In the past week, the inflow of US ETFs has remained above $40 billion, and the inflow of domestic stock products has exceeded $30 billion [2][13]. - The S&P 500 ETF of BlackRock continued to have the largest outflow, while the products of Vanguard had a large - scale inflow of over $40 billion, with a capital flow difference of over $80 billion between the two. The Russell 2000 and high - yield bond ETFs had inflows [2][15]. 2.2 US ETF Performance: Stock Long - Short and Other Alternative Strategies Perform Well - Many stock long - short products were issued last week, and products combining futures replication and multiple hedge fund strategies have been increasing in the past two years. Among the top ten alternative strategy products in the US, the multi - strategy product of State Street and the stock long - short product of Convergence performed the best [2][19]. 3. Recent Capital Flows of US Ordinary Mutual Funds - In October 2025, the total amount of US non - money mutual funds was $23.7 trillion, an increase of $0.22 trillion compared to September. The scale of domestic stock products increased by 0.9%, but the redemption pressure was still high [2][20]. - From November 25th to December 3rd, the outflow of US domestic stock funds remained above $15 billion. Hybrid products have continued to experience outflows recently, while bond funds have seen a slight inflow [2][20].

- The report discusses multiple quantitative models for market timing, including short-term, medium-term, and long-term models. These models are constructed based on principles such as price-volume relationships, momentum, and calendar effects. The short-term models include the "Volume Model," "Feature Institutional Model," and "Feature Volume Model," while medium-term models include the "Limit-Up/Down Model" and "Up/Down Return Difference Model." The long-term model is the "Long-Term Momentum Model"[8][11][12][13] - The construction process of these models involves combining signals from different time horizons and strategies. For example, the "Volume Model" evaluates market activity through trading volume, while the "Momentum Model" focuses on price trends. The "Limit-Up/Down Model" identifies market sentiment by analyzing the frequency of limit-up and limit-down events. The "Up/Down Return Difference Model" measures the difference between upward and downward returns to gauge market direction[8][11][12] - The evaluation of these models suggests that combining signals from different models enhances robustness. For instance, some models are defensive, while others are aggressive, allowing for a balanced approach. The report emphasizes that simplicity in model design often leads to better generalization and performance[8][11][12] - Backtesting results for these models indicate varying levels of effectiveness. For example, the "Long-Term Momentum Model" is currently bullish, while the "Up/Down Return Difference Model" shows a positive outlook across all broad-based indices. The "Feature Institutional Model" is bullish, whereas the "Feature Volume Model" is bearish. The "Volume Model" remains neutral across all indices[11][12][13]

证券研究报告 | 金融工程 gszqdatemark 2025 12 11 年 月 日 2）中证 500 和中证 1000 指数成分股在行业分布上均衡配置，涉及的行 业众多，主动选股的难度较大，而量化选股的空间相对较大。中证 500 和 中证 1000 指数的成分股的分散度较高，量化模型则可利用这一广度优势， 通过多因子框架系统性挖掘 500-1000 只成分股的定价偏差。 3）从市场中的 300、500、1000 指增产品来看，量化策略在中小盘指 数成分股内具备更高的获取超额的能力，即长期而言中小盘具备更高的 alpha 价值。相比大盘而言中小盘是更适合量化策略的土壤，指增产品可 以获取更加丰厚的 alpha 潜在回报。 基金 alpha 分析：竞争力与超额收益能力 。 1）历史业绩：大幅跑赢基准。兴银中证 500 指数增强 A 近 1 年展现出优 秀的增强实效：收益端实现 6.6%的稳健超额，风险端回撤控制严于基准， 风险调整后收益性价比尤为突出。兴银中证 1000 指数增强 A 近 1 年展现 出优秀的增强策略执行力：收益端实现 9.7%的显著超额，产品自成立以 来跑赢比较基准，显示出基金经理的量化选股 ...

Core Insights - The article highlights the performance and strategies of Zhixin Rongke, a quantitative private equity firm, which has shown impressive returns in the market, particularly in the CTA (Commodity Trading Advisor) category [4][13][24]. Company Overview - Zhixin Rongke Investment Management (Beijing) Co., Ltd. was established in 2013 by PhDs from Tsinghua University and the Chinese University of Hong Kong, focusing on quantitative investment with over 10 years of experience in CTA strategies and 5 years in quantitative stock strategies [13][14]. - The firm has developed a dual-driven strategy system centered on quantitative CTA and quantitative stock selection, aiming for high Sharpe ratios and low drawdowns [13][24]. Performance Metrics - As of October 2025, Zhixin Rongke's products have achieved significant average returns, ranking second among quantitative private equity firms and sixth among those with assets over 5 billion [4][10]. - The "Zhixin Rongke CTA No. 7 A Class" product ranked third in terms of returns and drawdown control among CTA products, showcasing its strong performance [4][8]. Investment Strategies - The firm employs a dual-engine strategy that combines CTA and quantitative stock selection, providing both trend-following returns and tail risk hedging [41][45]. - The strategies have demonstrated crisis alpha, achieving positive returns during market downturns, such as a +***% return when the CSI 300 index fell by 21.6% in 2022 [41][42]. Team and Development - The core team has over 15 years of stable collaboration, previously working at the renowned hedge fund WorldQuant, which enhances their research and investment capabilities [17][21]. - The firm has undergone several strategy iterations since its inception, continuously adapting to market changes and improving performance metrics [46]. Product Lines - Zhixin Rongke offers various product lines, including CTA-enhanced strategies and quantitative stock selection strategies, catering to different investor risk preferences [24][30]. - The "CTA No. 7" product is positioned as a flagship quantitative CTA product, while the "Multi-Strategy No. 8" integrates both CTA and quantitative stock selection for enhanced absolute returns [28][30].

（原标题：从逆风开局到领涨市场，兴银富利兴易智享量化实现40%净值跃升） 南财理财通数据显示，截至2025年11月24日，理财公司合计存续 549只1-3个月（含）期限的混合类公 募理财产品，近两年的平均净值涨幅为 4.34%，增长排名前十的产品分别来自信银理财、工银理财等4 家理财公司。值得注意的是，在理财公司混合类公募理财产品近两年业绩榜单（1-3月期限）中，兴银 理财占据五席，兴银理财"富利兴易智享量化指增3个月最短持有期1号混合类理财产品A"近两年净值增 长率达到 41.97%，收益断崖式领先，斩获榜单冠军。 2024年9月下旬起，美联储货币政策正式转向叠加国内政策超预期加码，市场热情被点燃，2025年三季 度A股市场更是成为全球亮点，主要指数全线上涨，科技板块领涨，整体呈现结构性牛市，产品紧跟权 益行情，全力打开上行弹性，净值节节攀升。截至2025年11月末，该产品成立以来净值增长率达到 40.42%，远超同期业绩比较基准，成立以来年化收益率也达到19.65%。 总结来看，兴银理财"富利兴易智享量化指增"系列依托其团队在量化策略研发方面的深厚专业素养和实 践经验，充分发挥了对市场数据的快速捕捉与分析 ...

Core Insights - The report emphasizes the construction of an AI-based low-frequency index enhancement strategy using deep learning frameworks to build alpha and risk models [3][14]. - The performance of various index enhancement funds has been highlighted, showing significant excess returns compared to their respective indices [10][11]. Market Index Performance - As of December 5, 2025, the Shanghai Composite Index rose by 0.37%, the Shenzhen Component Index increased by 1.26%, and the CSI 300 Index gained 1.28% [7][8]. - The year-to-date performance shows the CSI 300 Index up by 16.5%, while the CSI 300 index enhancement portfolio increased by 26.2%, resulting in an excess return of 9.7% [18]. Index Enhancement Fund Performance - For the CSI 300 index enhancement fund, the minimum excess return was -1.28%, the median was 0.11%, and the maximum was 0.95% for the week ending December 5, 2025 [10][11]. - Year-to-date, the CSI 500 index enhancement fund showed a minimum excess return of -10.18%, a median of 3.15%, and a maximum of 13.55% [11]. Tracking Portfolio Performance - The report outlines the construction of enhancement portfolios for the CSI 300, CSI 500, and CSI 1000 indices, utilizing deep learning to optimize alpha and risk signals [14][15]. - The CSI 500 index enhancement portfolio has achieved a year-to-date return of 30.3%, outperforming the CSI 500 index, which rose by 24.0%, resulting in an excess return of 6.4% [23][24]. Specific Index Enhancement Performance - The CSI A500 index enhancement portfolio has increased by 28.4% year-to-date, compared to a 19.6% rise in the CSI A500 index, yielding an excess return of 8.7% [29][32]. - The CSI 1000 index enhancement portfolio has shown a year-to-date increase of 38.0%, significantly outperforming the CSI 1000 index, which rose by 23.2%, leading to an excess return of 14.8% [35][36].