Quantitative Models and Factor Construction Quantitative Models and Construction Methods 1. Model Name: Broker Gold Stock Performance Enhancement Portfolio - **Model Construction Idea**: The model aims to optimize the selection from the broker gold stock pool to outperform the benchmark index of equity-biased hybrid funds[12][39] - **Model Construction Process**: - The model uses the broker gold stock pool as the stock selection space and constraint benchmark - It employs portfolio optimization to control deviations in individual stocks and styles from the broker gold stock pool - The industry allocation is based on the industry distribution of all public funds - The portfolio is adjusted at the closing price on the first day of each month[12][39][42] - **Model Evaluation**: The model has shown stable performance historically, consistently outperforming the equity-biased hybrid fund index annually from 2018 to 2022[12][39][42] Model Backtest Results Broker Gold Stock Performance Enhancement Portfolio - **Absolute Return (Monthly)**: -0.77% (20251009-20251031)[41] - **Excess Return Relative to Equity-biased Hybrid Fund Index (Monthly)**: 1.37% (20251009-20251031)[41] - **Absolute Return (Year-to-date)**: 35.08% (20250102-20251031)[41] - **Excess Return Relative to Equity-biased Hybrid Fund Index (Year-to-date)**: 2.61% (20250102-20251031)[41] - **Ranking in Active Equity Funds (Year-to-date)**: 40.13% percentile (412/3469)[41] Quantitative Factors and Construction Methods 1. Factor Name: Total Market Value - **Factor Construction Idea**: This factor measures the total market capitalization of a stock, which is often used to capture the size effect in stock returns[3][28] - **Factor Construction Process**: - The total market value is calculated as the product of the stock's current price and the total number of outstanding shares[3][28] - **Factor Evaluation**: The total market value factor has shown good performance in the recent month and year-to-date periods[3][28] 2. Factor Name: Single Quarter Revenue Growth Rate - **Factor Construction Idea**: This factor measures the growth rate of a company's revenue in a single quarter, indicating its short-term growth potential[3][28] - **Factor Construction Process**: - The single quarter revenue growth rate is calculated as the percentage change in revenue from the previous quarter to the current quarter[3][28] - **Factor Evaluation**: The single quarter revenue growth rate factor has shown good performance year-to-date[3][28] 3. Factor Name: Analyst Net Upward Revision - **Factor Construction Idea**: This factor measures the net number of upward revisions by analysts, reflecting positive changes in analyst sentiment[3][28] - **Factor Construction Process**: - The analyst net upward revision is calculated as the difference between the number of upward revisions and the number of downward revisions over a specific period[3][28] - **Factor Evaluation**: The analyst net upward revision factor has shown good performance year-to-date[3][28] Factor Backtest Results Total Market Value Factor - **Recent Month Performance**: Good[3][28] - **Year-to-date Performance**: Good[3][28] Single Quarter Revenue Growth Rate Factor - **Recent Month Performance**: Not specified - **Year-to-date Performance**: Good[3][28] Analyst Net Upward Revision Factor - **Recent Month Performance**: Not specified - **Year-to-date Performance**: Good[3][28]

Core Viewpoint - The article emphasizes the rapid development of AI and quantitative technology in the investment sector, highlighting the importance of continuous strategy evolution for the long-term success of quantitative private equity firms like Tianyan Capital, which was founded in 2014 and has a strong focus on innovation and adaptation [2]. Company Overview - Tianyan Capital was co-founded by Xie Xiaoyang and Zhang Sen, both of whom have over ten years of industry experience. The company’s name reflects its commitment to change and deep insights into the essence of investment [2]. - The firm has received multiple industry awards, including the "Golden Changjiang Award" and "Yinghua Award," and ranks among the top ten quantitative private equity firms in terms of performance [3][4]. Performance Metrics - As of September 2025, Tianyan Capital's products have achieved impressive returns, with an average return of ***% over the past three years, placing it first in the industry [3][4]. - The firm manages approximately 2.1 billion yuan across 11 products that meet ranking criteria, showcasing its strong long-term performance [3]. Investment Strategy - The core strategy of Tianyan Capital is centered around a multi-factor model for stock selection, which allows for higher alpha returns at a lower cost [8]. - The flagship product, "Tianyan Saineng," has been operational since May 2016 and has demonstrated significant returns, with a focus on maintaining model autonomy and stability in risk control [10][11]. Team and Culture - The investment research team at Tianyan Capital consists of over half PhD holders from prestigious institutions, fostering a culture of free exploration and innovation [12]. - The company emphasizes long-termism in its operations, avoiding arbitrary changes to risk parameters and maintaining a stable risk control model [10][11]. Market Position and Future Outlook - Tianyan Capital has strategically positioned itself to balance scale and performance, understanding that growth in assets under management should align with long-term performance and research capabilities [14]. - The firm has also obtained a Hong Kong license to enhance its global asset allocation capabilities, focusing on capturing unique alpha opportunities in the Chinese market while catering to international investors [16].

Report Industry Investment Rating - Not provided in the given content Core Viewpoints - Based on the Huatai Commodity Multi-Factor Model, this week it is recommended to overweight copper, silver, soybean oil, gold, and fresh apples, and underweight glass, alumina, soda ash, eggs, and styrene [4][51] Summary by Relevant Catalogs 1. Plate Liquidity - This week, the trading volume of the basic metals sector was 1784.354 billion yuan, a change of 104.21% from last week, with a margin of 50.724 billion yuan, a change of -3.33 billion yuan from last week [1] - The energy and chemical sector had a trading volume of 1641.153 billion yuan, a change of 148.50% from last week, and a margin of 36.5 billion yuan, a change of 0.198 billion yuan from last week [1] - The agricultural products sector had a trading volume of 1222.184 billion yuan, a change of 88.30% from last week, and a margin of 41.853 billion yuan, a change of 1.864 billion yuan from last week [1] - The precious metals sector had a trading volume of 5172.317 billion yuan, a change of 271.03% from last week, and a margin of 76.338 billion yuan, a change of 4.96 billion yuan from last week [1] - The black building materials sector had a trading volume of 1013.342 billion yuan, a change of 161.66% from last week, and a margin of 33.353 billion yuan, a change of 1.948 billion yuan from last week [1] - The stock index futures sector had a trading volume of 3921.85 billion yuan, a change of 133.22% from last week, and a margin of 154.917 billion yuan, a change of -10.672 billion yuan from last week [1] - The treasury bond futures sector had a trading volume of 1592.895 billion yuan, a change of 132.22% from last week, and a margin of 16.084 billion yuan, a change of 1.145 billion yuan from last week [1] 2. Market and Plate Style - Since the beginning of this year, the Wande Commodity Index has a change of 33.76%, the Non-ferrous Index has a change of 2.25%, the Energy Index has a change of -22.63%, the Chemical Index has a change of -17.92%, the Oilseeds Index has a change of 4.47%, the Precious Metals Index has a change of 48.17%, and the Coking Coal and Steel Ore Index has a change of 0.64% [2] - The Huatai Commodity Long-term Momentum Index has a change of 18.76%, the Short-term Momentum Index has a change of 0.20%, the Skewness Index has a change of 12.23%, and the Term Structure Index has a change of 3.39% [2] - The latest VIX indicators of stock index options are as follows: SSE 50 Index Option is 19.26%, CSI 300 Index Option is 20.98%, and CSI 1000 Index Option is 26.67% [2] 3. Plate Premium and Discount Structure - The latest basis of stock index futures: IH is 7.47 points, IF is -17.27 points, IC is -143.47 points, and IM is -159.17 points; the annualized basis rate: IH is 1.46%, IF is -2.22%, IC is -11.85%, and IM is -12.83% [3] - The latest basis of treasury bond futures: TS is -0.02 yuan, TF is -0.05 yuan, T is 0.10 yuan, and TL is -0.29 yuan; the latest net basis: TS is -0.01 yuan, TF is -0.04 yuan, T is -0.08 yuan, and TL is -0.51 yuan [3] 4. Strategy - According to the Huatai Commodity Multi-Factor Model, this week it is recommended to overweight copper, silver, soybean oil, gold, and fresh apples, and underweight glass, alumina, soda ash, eggs, and styrene [4][51]

- **Barra style factors**: The report tracks various style factors including Beta, Market Cap, Momentum, Volatility, Non-linear Market Cap, Valuation, Liquidity, Profitability, Growth, and Leverage. Each factor is constructed using specific financial metrics and formulas. For example, the Profitability factor combines analyst forecast earnings price ratio, inverse price-to-cash flow ratio, and inverse price-to-earnings ratio (TTM), among others. The Growth factor incorporates earnings growth rate and revenue growth rate. These factors are used to evaluate stocks based on their historical and financial characteristics [13][14][15]. - **GRU factors**: GRU factors are derived from different training objectives, such as predicting future one-day close-to-close or open-to-open returns. Examples include `close1d`, `open1d`, `barra1d`, and `barra5d`. These factors are constructed using GRU models trained on historical data to forecast short-term stock movements. GRU factors showed strong performance, with most models achieving positive multi-period returns, except for `barra1d`, which experienced some drawdowns [20][28][32]. - **Factor testing methodology**: Factors are tested using a long-short portfolio approach. At the end of each month, stocks are ranked based on the latest factor values, with the top 10% being long positions and the bottom 10% being short positions. The portfolios are equally weighted, and factors are industry-neutralized before testing. This methodology ensures robust evaluation of factor performance across different market conditions [15][16][31]. - **Factor performance results**: - **Style factors**: Valuation, Profitability, and Leverage factors showed strong long performance, while Beta, Liquidity, and Momentum factors performed well on the short side [15][16]. - **Technical factors**: Across various time windows, low momentum and low volatility stocks generally outperformed, while high volatility and high momentum stocks underperformed. For example, the 60-day momentum factor showed a negative return of -3.11% in the last month but a positive return of 2.12% over the last six months [19][26][30]. - **GRU factors**: GRU models like `barra1d` achieved a year-to-date excess return of 5.22%, while `barra5d` and `open1d` also delivered strong multi-period returns. However, `barra1d` experienced a weekly drawdown of -1.65% [20][32][33]. - **Multi-factor portfolio performance**: The multi-factor portfolio outperformed the benchmark (CSI 1000 Index) by 1.35% over the past week. GRU-based models also showed strong excess returns, ranging from 0.68% to 1.60% over the same period. Year-to-date, the `barra1d` model achieved an excess return of 5.22% [32][33][34].

Core Viewpoint - The article discusses the development and capabilities of the GF Quantitative Alpha Factor Database, which supports various investment strategies through a comprehensive factor library built on extensive research and data accumulation by the GF Quantitative team [1]. Group 1: Database Overview - The GF Quantitative Alpha Factor Database is established on MySQL 8.0 and encompasses over a decade of research experience, integrating fundamental factors, Level-1 and Level-2 high-frequency factors, machine learning factors, and alternative data factors [1]. - The database supports strategies such as long-short strategies, index enhancement, ETF rotation, asset allocation, and derivatives [1]. - The GF Quantitative team possesses a data storage capacity of over 100TB and high-performance CPU/GPU computing servers, collaborating with reliable data providers like Wind, Tianruan, and Tonglian for efficient factor development and dynamic updates [1]. Group 2: Factor Types and Performance - The article lists various factors categorized by type, including deep learning factors, trading volume factors, and market order ratios, each with specific definitions and performance metrics [3]. - For instance, the "agr_dailyquote" factor has a historical average of 14.22% and a historical win rate of 91.97% [3]. - The "bigbuy" factor shows a historical average of 7.85% with a win rate of 66.74% [3]. Group 3: Research Reports - A series of research reports are available for download, covering topics such as style factor-driven quantitative stock selection, industry selection, and macroeconomic observations related to Alpha factor trends [4][5]. - The reports include analyses on the application of factors in the CSI 300 index and various strategies for capturing industry alpha drivers [4].

Quantitative Models and Construction Methods Model Name: Broker Gold Stock Performance Enhancement Portfolio - **Model Construction Idea**: The model aims to optimize the selection of stocks from the broker gold stock pool to outperform the median of active equity funds[39][43]. - **Model Construction Process**: - The model uses the broker gold stock pool as the stock selection space and constraint benchmark. - It employs portfolio optimization to control deviations in individual stocks and styles from the broker gold stock pool. - The industry distribution of all public funds is used as the industry allocation benchmark. - The model's detailed construction method can be found in the report "Broker Gold Stock Full Analysis - Data, Modeling, and Practice" published on February 18, 2022[39][43]. - **Model Evaluation**: The model has shown stable performance historically, consistently outperforming the active equity fund index from 2018 to 2022, ranking in the top 30% of active equity funds each year[12][39]. Model Backtesting Results Broker Gold Stock Performance Enhancement Portfolio - **Absolute Return (Monthly)**: 15.49%[5][42] - **Excess Return Relative to Active Equity Fund Index (Monthly)**: 3.59%[5][42] - **Absolute Return (Year-to-Date)**: 34.01%[5][42] - **Excess Return Relative to Active Equity Fund Index (Year-to-Date)**: 5.72%[5][42] - **Ranking in Active Equity Funds (Year-to-Date)**: 30.38% percentile (1054/3469)[5][42] Quantitative Factors and Construction Methods Factor Name: Single Quarter Net Profit Growth Rate - **Factor Construction Idea**: This factor measures the growth rate of net profit in a single quarter[3][28]. - **Factor Construction Process**: - Calculate the net profit for the current quarter. - Compare it to the net profit of the same quarter in the previous year. - The formula is: $ \text{Net Profit Growth Rate} = \frac{\text{Current Quarter Net Profit} - \text{Previous Year Same Quarter Net Profit}}{\text{Previous Year Same Quarter Net Profit}} \times 100\% $ - **Factor Evaluation**: This factor has performed well recently[3][28]. Factor Name: Single Quarter ROE - **Factor Construction Idea**: This factor measures the return on equity for a single quarter[3][28]. - **Factor Construction Process**: - Calculate the net income for the quarter. - Divide it by the average shareholders' equity for the quarter. - The formula is: $ \text{ROE} = \frac{\text{Net Income}}{\text{Average Shareholders' Equity}} \times 100\% $ - **Factor Evaluation**: This factor has performed well recently[3][28]. Factor Name: Single Quarter Revenue Growth Rate - **Factor Construction Idea**: This factor measures the growth rate of revenue in a single quarter[3][28]. - **Factor Construction Process**: - Calculate the revenue for the current quarter. - Compare it to the revenue of the same quarter in the previous year. - The formula is: $ \text{Revenue Growth Rate} = \frac{\text{Current Quarter Revenue} - \text{Previous Year Same Quarter Revenue}}{\text{Previous Year Same Quarter Revenue}} \times 100\% $ - **Factor Evaluation**: This factor has performed well recently[3][28]. Factor Backtesting Results Single Quarter Net Profit Growth Rate - **Performance**: This factor has shown good performance recently[3][28]. Single Quarter ROE - **Performance**: This factor has shown good performance recently[3][28]. Single Quarter Revenue Growth Rate - **Performance**: This factor has shown good performance recently[3][28].

Core Insights - The article discusses the use of a multi-factor model suitable for the CSI 300 index constituents, combined with a small-cap high-growth satellite strategy, to enhance the performance of the CSI 300 enhanced strategy [1][2] - Since 2016, the CSI 300 enhanced strategy has achieved an annualized excess return of 12.6% with a tracking error of 5.2% under a satellite allocation of 30% domestic and 10% foreign [1][2] Summary by Sections - **Performance Analysis**: The CSI 300 enhanced strategy has shown an annualized excess return of at least 10% since 2016, with an information ratio exceeding 2.0. The internal component of the strategy has lower tracking error and relative drawdown, while the external component offers greater return elasticity but with higher tracking error and drawdown [1][2] - **Model Construction**: The multi-factor model is constructed based on fundamental and momentum indicators, which has demonstrated better stock selection robustness compared to the all-A multi-factor model [1] - **Satellite Strategy**: The external component can be replaced with small-cap high-growth or GARP strategies. The optimal satellite allocation depends on the risk-return preference, with the most extreme case showing an annualized excess return of 17.5% when fully utilizing satellite strategies [2]

- The report discusses a multi-factor model suitable for the constituents of the CSI 300 Index, combined with a small-cap high-growth portfolio as an external satellite strategy to improve the performance of the CSI 300 enhanced strategy[1][3][5] - The internal part of the enhanced strategy uses a multi-factor model based on fundamental and momentum indicators, including factors such as ROE, ROE YoY, SUE, expected net profit adjustment, accelerated growth, cash flow ratio, value (dividend yield and BP equal weight composite), momentum, buy-in strength after opening, and large order-driven rise[16][17] - The external part of the enhanced strategy uses a small-cap high-growth portfolio, constructed using factors such as SUE, EAV, expected net profit adjustment, cumulative R&D investment, PB_INT, small-cap, late trading volume ratio, and large order net buy-in ratio after opening[35][36] - The internal multi-factor model shows more stable stock selection performance within the CSI 300 Index constituents compared to the all-A multi-factor model, with higher IC and RankIC information ratios[16][17] - The small-cap high-growth portfolio has an annualized return of 25.0% since 2016, with an annualized excess return of 24.4% relative to the CSI 300 Index, but also higher tracking error[35][36] - The GARP strategy, which balances growth potential and reasonable pricing, is also considered as an external satellite strategy, showing an annualized return of 20.9% for the GARP 20 portfolio and 17.4% for the GARP 50 portfolio since 2016[39][40][42] - Combining the internal multi-factor model and external satellite strategies (small-cap high-growth or GARP) can significantly improve the performance of the CSI 300 enhanced strategy, with annualized excess returns not less than 10% and information ratios above 2.0 since 2016[29][45][55] Model and Factor Construction Process - **Internal Multi-Factor Model**: Constructed using fundamental and momentum indicators, including ROE, ROE YoY, SUE, expected net profit adjustment, accelerated growth, cash flow ratio, value (dividend yield and BP equal weight composite), momentum, buy-in strength after opening, and large order-driven rise[16][17] - **Small-Cap High-Growth Portfolio**: Constructed using factors such as SUE, EAV, expected net profit adjustment, cumulative R&D investment, PB_INT, small-cap, late trading volume ratio, and large order net buy-in ratio after opening[35][36] - **GARP Strategy**: Constructed by excluding high-risk stocks, using PB and dividend yield as value factors, and ROE, SUE, EAV, expected net profit adjustment, and two-year compound growth rate as growth factors, selecting the top 20 or 50 stocks based on composite scores[41][42] Model and Factor Performance Metrics - **Internal Multi-Factor Model**: IC monthly average 6.36%, IC monthly win rate 67.0%, annualized ICIR 1.67; RankIC monthly average 7.53%, RankIC monthly win rate 72.2%, annualized ICIR 2.00[17] - **Small-Cap High-Growth Portfolio**: Annualized return 25.0%, annualized excess return 24.4%, tracking error 20.3%, information ratio 1.21, relative drawdown 39.6%, monthly win rate 61.4%[36] - **GARP 20 Portfolio**: Annualized return 20.9%, annualized excess return 20.3%, tracking error 15.8%, information ratio 1.26, relative drawdown 36.0%[42] - **GARP 50 Portfolio**: Annualized return 17.4%, annualized excess return 16.8%, tracking error 14.6%, information ratio 1.14, relative drawdown 37.2%[42] Combined Strategy Performance - **Internal 20% + External 10% (Small-Cap High-Growth)**: Annualized excess return 11.7%, information ratio 2.35, tracking error 5.2%, relative drawdown 21.9%[45][48] - **Internal 20% + External 10% (GARP)**: Annualized excess return 11.3%, information ratio 2.41, tracking error 4.3%, relative drawdown 5.8%[50][53]

Quantitative Models and Construction Methods Model Name: Hong Kong Stock Selection 20 Portfolio - **Model Construction Idea**: The model aims to track the monthly performance of a long portfolio by selecting the top 20 stocks with the highest scores based on multiple factors. The benchmark is the Hong Kong Composite Index (HKD) (930930.CSI) [40][42] - **Model Construction Process**: - The model uses four types of factors: technical, capital, fundamental, and analyst expectations [40][41] - At the end of each month, the top 20 stocks with the highest scores are selected and equally weighted to construct the Hong Kong Stock Selection 20 Portfolio [42] - **Model Evaluation**: The model has shown excellent performance in the Hong Kong Stock Connect sample stocks [40] Model Backtesting Results - **Hong Kong Stock Selection 20 Portfolio**: - **July 2025**: Portfolio return 11.6%, benchmark index return 4.8%, excess return 6.8% [42] - **Overall Period (2015.1~2025.7)**: Excess annualized return 13.9%, excess return volatility ratio 1.0 [42][44] Quantitative Factors and Construction Methods Factor Name: Technical, Capital, Fundamental, Analyst Expectations - **Factor Construction Idea**: These factors are used to evaluate the performance of Hong Kong Stock Connect sample stocks [40] - **Factor Construction Process**: - **Technical**: Based on stock price movements and technical indicators - **Capital**: Based on capital flow and trading volume - **Fundamental**: Based on financial metrics such as PE ratio, ROE, etc. - **Analyst Expectations**: Based on analyst ratings and forecasts [40][41] - **Factor Evaluation**: These factors have shown excellent performance in the Hong Kong Stock Connect sample stocks [40] Factor Backtesting Results - **Technical, Capital, Fundamental, Analyst Expectations**: - **July 2025**: Excess annualized return 13.9%, excess return volatility ratio 1.0 [42][44]

Group 1: Report Industry Investment Rating - No information provided in the content Group 2: Core Viewpoints of the Report - Optimizing the position strategy can effectively enhance the real - world performance of the quantitative framework, which is a multiplier method for increasing returns, especially in volatile markets [2][6][111] - Binary full - position strategies can capture returns efficiently in obvious trends but come with high volatility, drawdown risks, and high turnover and commission costs; threshold step - by - step addition strategies have low trading frequency but limited ability to capture returns in volatile markets (except for the LG model) [2][111] - Single continuous strategies perform well in volatile markets. Linear and normal strategies show high return stability, while Sigmoid, Atanh, and Atanh - Sigmoid strategies have significant advantages in volatility control, suitable for risk - averse investors. The GRU model shows stable performance in improving odds, while the strategy advantages of LG, SVM and other models are environment - dependent [2][111] - In terms of turnover and commission consumption, single continuous strategies such as Sigmoid and Atanh can reduce turnover and commission consumption in volatile markets, and investors should focus more on returns rather than commission costs [2][111] Group 3: Summary According to the Table of Contents 3.1 Multi - factor Model's Position Strategy Introduction - **Multi - long and short full - position strategy**: It is a binary extreme position management mode, which can be used as a performance benchmark and a reference for other strategies. It performs poorly in bull markets and better in volatile markets, and is more suitable for non - linear models in volatile markets [12][32][33] - **Threshold multi - long and short full - position strategy and step - by - step addition strategy**: The threshold full - position strategy introduces a fuzzy interval filtering mechanism to reduce misjudgment risks and improve the overall risk - return ratio. The step - by - step addition strategy can reduce turnover and trading costs but may sacrifice some returns, except for the LG model in volatile markets [13][14][53] - **Continuous strategies based on different risk preferences and mapping functions**: Continuous strategies can convert binary probability signals into position adjustment signals, which can be divided into risk - seeking, risk - averse, and risk - neutral types according to risk preferences. Different mapping functions such as linear, Sigmoid, normal, Atanh, and Atanh - Sigmoid are used [18] 3.2 Strategy Back - testing - **Back - testing sample interval and key parameters**: The trading target is the Treasury bond futures T contract. The period from January 1, 2024, to December 31, 2024, is regarded as a bull market, and the period from January 1, 2025, to May 9, 2025, is regarded as a volatile market [31] - **Benchmark results of multi - long and short full - position strategy**: It has little effect on increasing returns in bull markets and performs better in volatile markets. Non - linear models such as RF and SVM can better handle the problem of return increase in volatile markets [33][34] - **Threshold full - position strategy and step - by - step position adjustment strategy**: The threshold strategy can optimize the odds of investment strategies in both bull and volatile markets, but the application effect depends on the model type and market environment. The step - by - step position adjustment strategy can significantly reduce turnover and trading costs but usually sacrifices some returns, except for the LG model in volatile markets [37][40][53] - **Analysis of the effect of single continuous strategies**: In volatile markets, continuous position strategies can significantly improve the odds of strategies without increasing the prediction win rate. Different strategies such as Atanh and Sigmoid have different risk - return characteristics, and their turnover is related to the model and market environment [58][73][74] - **Rediscussion of the impact of trading commissions**: The key is to increase returns rather than reduce costs. Although different models and strategies have different commission consumption, the impact of commissions on returns is relatively small, and investors should focus on return enhancement [94][97][110] 3.3 Summary and Strategy Recommendations - Different position management strategies play an important role in return acquisition and risk control. Investors should choose appropriate models and strategies according to their risk preferences and market conditions [111]