Quantitative Model and Construction 1. Model Name: Coal Index Trend Tracking Model - **Model Construction Idea**: The model assumes that the price movement of the target has strong local continuity, with prices always in a certain trend. Reversal periods are significantly shorter than trend continuation periods. In cases of narrow consolidation, the model assumes the continuation of the previous trend. For large-scale trends, given a short observation window, the movement will follow the local trend within the window. When a reversal occurs, the price change at the start and end of the observation window will exceed the range caused by random fluctuations, thus filtering out random noise[2][3]. - **Model Construction Process**: 1. Calculate the difference between the closing price on day T and day T-20, denoted as `del` 2. Calculate the volatility (`Vol`) over the period from T-20 to T (exclusive) 3. If the absolute value of `del` exceeds N times `Vol`, the current price is considered to have broken out of the original oscillation range, forming a trend. The trend direction (long/short) corresponds to the sign of `del`. Otherwise, the trend direction remains the same as on day T-1 4. For tracking, N is set to 1, considering the higher volatility of the stock market compared to the bond market, which provides more short-term opportunities 5. The model evaluates both long and short returns for the coal index and combines the results for final assessment[3] - **Model Evaluation**: The model is not suitable for direct application to the Shenwan Level-1 Coal Index due to poor annualized return performance and significant drawdowns during the tracking period[4] --- Model Backtesting Results 1. Coal Index Trend Tracking Model - **Annualized Return**: -8.01%[3] - **Annualized Volatility**: 22.67%[3] - **Sharpe Ratio**: -0.35[3] - **Maximum Drawdown**: 22.79%[3] - **Total Return**: -8.75%[3]

2025 年 05 月 18 日 或仅是整理，蓄势以待机 本期要点：或仅是整理，蓄势以待机 上期提到，全天候模型发出了技术面上的风险提示信号，从这个角度 看未来一段时间市场或将进入技术上的震荡整理期。事后来看，走势 基本符合预期。 对于大盘，从走势形态看，本轮反弹已经涨了两波，第二波是从五一 之后开始的。从周期分析模型对不同级别趋势的监控分析结果来看， 当前潜在的调整仅是针对于五一之后上涨过程的调整，而四月初以来 的上行趋势或并没有已经结束的充分证据。从这个角度看，本轮调整 或最多回到 4 月底 5 月初的震荡区间，随后有望重新开始一波上行 走势，只有在那波上行趋势再次衰竭的时候才能判断本轮反弹是否真 正结束了。 在当前潜在的蓄势震荡过程中，四轮驱动模型建议关注军工、家电、 农林牧渔、汽车、电子、计算机、非银等板块。整体风格上可以适当 偏小盘，结构上可以适当偏均衡一些。 风险提示：根据历史数据构建的模型在市场变化时可能失效。 金融工程定期报告 证券研究报告 杨勇 分析师 SAC 执业证书编号：S1450518010002 yangyong1@essence.com.cn 相关报告 结构重于仓位，继续关注先 进 ...

Investment Rating - The industry is rated positively, with expectations of overall returns exceeding the CSI 300 Index by more than 5% over the next six months [13]. Core Insights - The model assumes that the price trends of the underlying assets exhibit strong local continuity, with reversal trends lasting significantly shorter than trend continuation periods. In cases of narrow consolidation, it is assumed that the previous trend will continue [4]. - The model's annualized return is reported at 24.17%, with a volatility of 26.54% and a Sharpe ratio of 0.91. The maximum drawdown recorded is 14.04%, and the total return rate during the index period is 19.46% [4]. - The model is designed for the Shenwan Level 1 Non-Bank Financial Index and is suitable for tracking this index from March 7, 2023, to March 18, 2025 [4][5]. Summary by Sections Model Overview - The model calculates the difference between the closing price on day T and the closing price 20 days prior, assessing volatility over the same period. A significant deviation from the historical volatility indicates a trend formation [4]. Performance Evaluation - The model's net value has shown a slow upward trend from March 7, 2023, to September 12, 2024, with a sharp increase influenced by market conditions and macro policies during a specific period [5]. - The model is deemed suitable for the Shenwan Level 1 Non-Bank Financial Index, demonstrating good annualized returns without prolonged significant drawdowns [5].

Investment Rating - The industry rating is "Neutral," indicating that the overall return is expected to be between -5% and 5% relative to the CSI 300 index over the next six months [12]. Core Insights - The model used for evaluating the defense and military industry assumes that price trends exhibit good local continuity, with reversals occurring less frequently than trend continuations. The model aims to identify trends based on price movements and volatility [4][5]. - The evaluation period for the model is from March 7, 2023, to March 18, 2025, with a focus on the Shenwan Level 1 Defense and Military Industry Index [4]. - The model's annualized return is -0.37%, with a volatility of 29.80% and a maximum drawdown of 29.79%. The total return during the evaluation period is -3.81% [4][5]. Summary by Sections Model Overview - The model is designed to track price movements and identify trends based on historical data, with specific algorithms to determine when a price has deviated from its previous range [4]. Results Assessment - The model experienced a decline in net value during specific periods, indicating a long-term drawdown and an inability to achieve favorable cumulative returns. The model is deemed unsuitable for direct application to the Shenwan Level 1 Defense and Military Industry Index due to negative returns [5]. Performance Metrics - The model's performance metrics include an annualized return of -0.37%, a volatility of 29.80%, a Sharpe ratio of -0.01, and a maximum drawdown of 29.79% [4].

金 金融工程点评 [Table_Message]2025-05-12 建筑材料指数趋势跟踪模型效果点评 [Table_Author] 证券分析师：刘晓锋 电话：13401163428 E-MAIL：liuxf@tpyzq.com 执业资格证书编码：S1190522090001 研究助理：孙弋轩 电话：18910596766 E-MAIL：sunyixuan@tpyzq.com 一般证券业务登记编码：S1190123080008 模型概述 结果评估： 区间年化收益：1.98% 波动率（年化）：24.36% 夏普率：0.08 最大回撤：25.11% 指数期间总回报率：-29.59% 太 平 洋 证 券 股 份 有 限 公 司 证 券 研 究 报 [Table_Title] [Table_Summary] 融 工 程 点 评 告 ◼ 设计原理：模型假定标的价格走势具有很好的局部延续性，标的价格永远处 于某一趋势中，出现反转行情的持续时间明显小于趋势延续的时间，若出现 窄幅盘整的情况，亦假设其延续之前的趋势。当处于大级别的趋势之中时， 给定较短时间的观察窗口，走势将延续观察窗口内的局部趋势。而当趋势发 生反转时，在观 ...

Quantitative Factors and Construction Methods Factor Name: Beta - Construction Idea: Measures the sensitivity of a stock's returns to market returns[14] - Construction Process: Calculated using the weighted regression of individual stock returns against market returns over the last 490 trading days[14] - Evaluation: Beta factor performed well in the recent week[8][10] Factor Name: Turnover Rate and Average Price Correlation (1 Month) - Construction Idea: Measures the correlation between turnover rate and average price over the past month[13] - Construction Process: Calculated as the correlation coefficient between turnover rate and average price over the past 20 trading days[13] - Evaluation: This factor showed good performance in the recent week and month[8][10] Factor Name: Turnover Rate Volatility (1 Month) - Construction Idea: Measures the volatility of turnover rate over the past month[13] - Construction Process: Calculated as the standard deviation of turnover rate over the past 20 trading days[13] - Evaluation: This factor performed well in the recent month and year[8][10] Factor Name: Reversal (1 Month) - Construction Idea: Measures the cumulative returns over the past month[13] - Construction Process: Calculated as the cumulative returns over the past 20 trading days[13] - Evaluation: This factor showed good performance in the recent week and month[8][10] Factor Name: Specificity (1 Month) - Construction Idea: Measures the specificity of stock returns relative to the Fama-French three-factor model[13] - Construction Process: Calculated as 1 minus the R-squared value from the regression of daily returns against the Fama-French three factors over the past 20 trading days[13] - Evaluation: This factor performed well in the recent year[8][10] Factor Name: Residual Volatility (Fama-French Three-Factor Model, 1 Month) - Construction Idea: Measures the residual volatility of stock returns relative to the Fama-French three-factor model[13] - Construction Process: Calculated as the standard deviation of residuals from the regression of daily returns against the Fama-French three factors over the past 20 trading days[13] - Evaluation: This factor showed good performance in the recent year[8][10] Factor Name: Excess Return Volatility (1 Month) - Construction Idea: Measures the volatility of excess returns over the past month[13] - Construction Process: Calculated as the standard deviation of excess returns over the past 20 trading days[13] - Evaluation: This factor performed well in the recent year[8][10] Factor Name: Small Market Capitalization - Construction Idea: Measures the logarithm of market capitalization[13] - Construction Process: Calculated as the logarithm of market capitalization[13] - Evaluation: This factor showed good performance in the recent week and year[8][10] Factor Backtesting Results Information Coefficient (IC) Performance - Beta: Recent week IC: 13.69%, Recent month IC: 0.85%, Recent year IC: 1.73%, Historical IC: 0.44%[9] - Turnover Rate and Average Price Correlation (1 Month): Recent week IC: 11.30%, Recent month IC: 7.07%, Recent year IC: 2.49%, Historical IC: 1.70%[9] - Turnover Rate Volatility (1 Month): Recent week IC: 6.15%, Recent month IC: 5.29%, Recent year IC: 2.99%, Historical IC: 2.51%[9] - Reversal (1 Month): Recent week IC: 11.08%, Recent month IC: 4.52%, Recent year IC: 2.87%, Historical IC: 2.15%[9] - Specificity (1 Month): Recent week IC: 11.05%, Recent month IC: 3.76%, Recent year IC: 3.63%, Historical IC: 2.41%[9] - Residual Volatility (Fama-French Three-Factor Model, 1 Month): Recent week IC: 5.42%, Recent month IC: 3.27%, Recent year IC: 3.62%, Historical IC: 2.48%[9] - Excess Return Volatility (1 Month): Recent week IC: -0.20%, Recent month IC: 1.88%, Recent year IC: 3.29%, Historical IC: 2.18%[9] - Small Market Capitalization: Recent week IC: 7.12%, Recent month IC: 2.70%, Recent year IC: 2.03%, Historical IC: 1.89%[9] Long Portfolio Performance - Beta: Recent week excess return: 1.08%, Recent month excess return: -0.75%, Recent year excess return: 6.46%, Historical cumulative excess return: -5.34%[11] - Turnover Rate and Average Price Correlation (1 Month): Recent week excess return: 1.08%, Recent month excess return: 2.92%, Recent year excess return: 2.72%, Historical cumulative excess return: 16.63%[11] - Turnover Rate Volatility (1 Month): Recent week excess return: 0.93%, Recent month excess return: 1.96%, Recent year excess return: 10.68%, Historical cumulative excess return: 32.01%[11] - Reversal (1 Month): Recent week excess return: 0.37%, Recent month excess return: 0.22%, Recent year excess return: 0.75%, Historical cumulative excess return: -1.18%[11] - Specificity (1 Month): Recent week excess return: 0.67%, Recent month excess return: 0.67%, Recent year excess return: 10.17%, Historical cumulative excess return: 16.91%[11] - Residual Volatility (Fama-French Three-Factor Model, 1 Month): Recent week excess return: 0.34%, Recent month excess return: 0.82%, Recent year excess return: 8.10%, Historical cumulative excess return: 18.57%[11] - Excess Return Volatility (1 Month): Recent week excess return: 0.02%, Recent month excess return: 0.09%, Recent year excess return: 7.20%, Historical cumulative excess return: 10.83%[11] - Small Market Capitalization: Recent week excess return: 0.95%, Recent month excess return: 0.12%, Recent year excess return: 10.84%, Historical cumulative excess return: 59.20%[11]

Group 1 - The core viewpoint of the report is that the model assumes a cyclical pattern of price deviation and regression relative to a reference index, with a defined limit on the degree of deviation, allowing for strategic buying when prices approach this limit [4][5]. - The model's design principle involves statistical analysis of historical data to identify reasonable thresholds for price deviations, which can signal buying opportunities when prices fall below these thresholds [4][5]. - The model tracks the performance of the Shenwan Level 1 Defense Industry Index relative to the CSI 300 Index over a defined period from January 4, 2010, to March 18, 2025 [4][5]. Group 2 - The total return of the interval strategy is reported at 159.57%, significantly outperforming the buy-and-hold return of 42.53%, resulting in an excess return of 117.05% [4]. - The maximum drawdown recorded is 50.87%, with the longest drawdown period lasting 2108 trading days, indicating substantial volatility in the strategy's performance [4]. - The model's effectiveness is questioned due to the observed price movements falling outside the historical sample range, suggesting that the strategy may not provide reliable guidance for future investments in the defense industry index [5].

Group 1: ETF Rotation Strategy Tracking - The ETF rotation strategy, constructed using GBDT+NN machine learning factors, has shown excellent performance in out-of-sample testing, with an IC value of 20.91% and a long position excess return of 0.61% last week [2][12][13] - The strategy's annualized excess return is 11.91%, with a maximum drawdown of 17.31% and an information ratio of 0.68, indicating strong recent performance [2][18][16] - The strategy has recorded an excess return of 0.88% last week, 1.44% for the month, and 0.15% year-to-date, reflecting its recent success [2][18] Group 2: High-Frequency Factor Overview - Various high-frequency factors have demonstrated strong performance, with the price range factor achieving a long position excess return of 1.01% last week and 5.84% year-to-date [3][22] - The volume-price divergence factor has shown a long position excess return of 10.13% this year, while the regret avoidance factor has underperformed with a return of -0.30% [3][22] - The overall performance of high-frequency factors has been commendable, with the price range factor and volume-price divergence factor leading in returns [3][22] Group 3: High-Frequency Factor Performance Tracking - The price range factor measures the activity level of stocks within different price ranges, indicating investor expectations for future price movements, and has shown stable performance this year [4][25] - The volume-price divergence factor assesses the correlation between stock price and trading volume, with lower correlation suggesting higher future price increases, although its performance has been inconsistent in recent years [4][25] - The regret avoidance factor reflects investor behavior, showing stable excess returns, indicating that regret avoidance sentiment significantly impacts expected stock returns [4][25] Group 4: Combined Strategies Performance - The high-frequency "gold" combination strategy has an annualized excess return of 10.68% and a maximum drawdown of 6.04%, with recent excess returns of 0.14% last week and 5.98% year-to-date [5][54] - The high-frequency and fundamental resonance combination strategy has shown an annualized excess return of 14.98% and a maximum drawdown of 4.52%, with recent excess returns of 0.28% last week and 4.99% year-to-date [5][60]

Quantitative Factors and Construction Methods - **Factor Name**: bp **Construction Idea**: Measures valuation by comparing net assets to market value **Construction Process**: Calculated as: $ bp = \frac{\text{Current Net Assets}}{\text{Current Total Market Value}} $ [13] **Evaluation**: Commonly used valuation factor, straightforward and widely applicable [13] - **Factor Name**: bp three-year percentile **Construction Idea**: Tracks the relative valuation of a stock over the past three years **Construction Process**: Represents the percentile rank of the current bp value within the last three years [13] **Evaluation**: Useful for identifying stocks with consistent valuation trends [13] - **Factor Name**: Quarterly ep **Construction Idea**: Measures profitability relative to net assets **Construction Process**: Calculated as: $ \text{Quarterly ep} = \frac{\text{Quarterly Net Profit}}{\text{Net Assets}} $ [13] **Evaluation**: Reflects short-term profitability, sensitive to quarterly fluctuations [13] - **Factor Name**: Quarterly ep one-year percentile **Construction Idea**: Tracks the relative profitability of a stock over the past year **Construction Process**: Represents the percentile rank of the current quarterly ep value within the last year [13] **Evaluation**: Helps identify stocks with improving or declining profitability trends [13] - **Factor Name**: Quarterly sp **Construction Idea**: Measures revenue generation relative to net assets **Construction Process**: Calculated as: $ \text{Quarterly sp} = \frac{\text{Quarterly Revenue}}{\text{Net Assets}} $ [13] **Evaluation**: Indicates operational efficiency, useful for growth-oriented analysis [13] - **Factor Name**: Quarterly sp one-year percentile **Construction Idea**: Tracks the relative operational efficiency of a stock over the past year **Construction Process**: Represents the percentile rank of the current quarterly sp value within the last year [13] **Evaluation**: Highlights trends in revenue generation efficiency [13] - **Factor Name**: Quarterly asset turnover **Construction Idea**: Measures revenue generation relative to total assets **Construction Process**: Calculated as: $ \text{Quarterly Asset Turnover} = \frac{\text{Quarterly Revenue}}{\text{Total Assets}} $ [13] **Evaluation**: Reflects operational efficiency, sensitive to asset-heavy industries [13] - **Factor Name**: Quarterly gross margin **Construction Idea**: Measures profitability relative to sales revenue **Construction Process**: Calculated as: $ \text{Quarterly Gross Margin} = \frac{\text{Quarterly Gross Profit}}{\text{Quarterly Sales Revenue}} $ [13] **Evaluation**: Indicates pricing power and cost control [13] - **Factor Name**: Quarterly roa **Construction Idea**: Measures profitability relative to total assets **Construction Process**: Calculated as: $ \text{Quarterly ROA} = \frac{\text{Quarterly Net Profit}}{\text{Total Assets}} $ [13] **Evaluation**: Reflects overall asset efficiency [13] - **Factor Name**: Quarterly roe **Construction Idea**: Measures profitability relative to net assets **Construction Process**: Calculated as: $ \text{Quarterly ROE} = \frac{\text{Quarterly Net Profit}}{\text{Net Assets}} $ [13] **Evaluation**: Commonly used profitability metric, sensitive to leverage [13] - **Factor Name**: Standardized unexpected earnings **Construction Idea**: Measures deviation of current earnings from historical growth trends **Construction Process**: Calculated as: $ \text{Standardized Unexpected Earnings} = \frac{\text{Current Quarterly Net Profit} - (\text{Last Year Same Quarter Net Profit} + \text{Average Growth of Last 8 Quarters})}{\text{Standard Deviation of Growth in Last 8 Quarters}} $ [13] **Evaluation**: Useful for identifying earnings surprises [13] - **Factor Name**: Standardized unexpected revenue **Construction Idea**: Measures deviation of current revenue from historical growth trends **Construction Process**: Calculated as: $ \text{Standardized Unexpected Revenue} = \frac{\text{Current Quarterly Revenue} - (\text{Last Year Same Quarter Revenue} + \text{Average Growth of Last 8 Quarters})}{\text{Standard Deviation of Growth in Last 8 Quarters}} $ [13] **Evaluation**: Highlights revenue surprises [13] - **Factor Name**: Dividend yield **Construction Idea**: Measures dividend payout relative to market value **Construction Process**: Calculated as: $ \text{Dividend Yield} = \frac{\text{Annual Dividend}}{\text{Current Market Value}} $ [13] **Evaluation**: Commonly used for income-focused strategies [13] - **Factor Name**: 1-month turnover rate volatility **Construction Idea**: Measures the variability of turnover rates over the past month **Construction Process**: Calculated as the standard deviation of daily turnover rates over the past 20 trading days [13] **Evaluation**: Reflects liquidity and trading activity [13] - **Factor Name**: Fama-French three-factor residual volatility **Construction Idea**: Measures the volatility of residuals from a Fama-French three-factor model regression **Construction Process**: Calculated as the standard deviation of residuals from daily returns regressed on the Fama-French three factors over the past 20 trading days [13] **Evaluation**: Indicates idiosyncratic risk [13] Factor Backtesting Results - **Factor Name**: bp **IC Values**: Weekly: -8.41%, Monthly: 3.48%, Yearly: 1.72% [8] **Excess Return**: Weekly: -0.18%, Monthly: 1.03%, Yearly: 3.10% [11] - **Factor Name**: bp three-year percentile **IC Values**: Weekly: 2.04%, Monthly: 7.90%, Yearly: 2.82% [8] **Excess Return**: Weekly: 0.67%, Monthly: 0.82%, Yearly: 2.55% [11] - **Factor Name**: Quarterly ep **IC Values**: Weekly: -5.19%, Monthly: 3.65%, Yearly: 0.24% [8] **Excess Return**: Weekly: -1.30%, Monthly: -0.08%, Yearly: 1.51% [11] - **Factor Name**: Quarterly ep one-year percentile **IC Values**: Weekly: 1.73%, Monthly: 4.68%, Yearly: 1.02% [8] **Excess Return**: Weekly: -0.35%, Monthly: 0.93%, Yearly: 4.35% [11] - **Factor Name**: Quarterly sp **IC Values**: Weekly: -5.87%, Monthly: -1.49%, Yearly: 0.18% [8] **Excess Return**: Weekly: -0.28%, Monthly: -1.23%, Yearly: 0.24% [11] - **Factor Name**: Quarterly sp one-year percentile **IC Values**: Weekly: 1.93%, Monthly: 7.04%, Yearly: 2.70% [8] **Excess Return**: Weekly: -0.66%, Monthly: 0.55%, Yearly: 3.80% [11] - **Factor Name**: Standardized unexpected earnings **IC Values**: Weekly: 0.24%, Monthly: 2.19%, Yearly: 0.64% [8] **Excess Return**: Weekly: -0.60%, Monthly: -0.75%, Yearly: 3.99% [11] - **Factor Name**: Standardized unexpected revenue **IC Values**: Weekly: -1.03%, Monthly: 0.72%, Yearly: 0.61% [8] **Excess Return**: Weekly: -0.41%, Monthly: -0.72%, Yearly: 1.55% [11] - **Factor Name**: Dividend yield **IC Values**: Weekly: -2.91%, Monthly: 1.85%, Yearly: -0.07% [8] **Excess Return**: Weekly: -0.37%, Monthly: 1.27%, Yearly: -4.85% [11]

Investment Rating - The industry investment rating is "Outperform the Market," indicating an expected industry index increase of over 5% in the next six months [18]. Core Insights - Recent factor performance shows that the average turnover rate, non-liquid shock, and quarterly gross margin factors have performed well, while factors like Beta and one-year momentum have underperformed [2][9]. - Over the past year, small-cap stocks, earnings forecast accuracy, and one-month turnover rate volatility have shown strong performance, while one-year momentum and expected adjustment averages have lagged [2][9]. Factor Tracking Summary Factor IC Performance - In the last week, the one-month average turnover rate, non-liquid shock, and turnover rate volatility factors performed well, while one-year momentum and quarterly asset turnover rate showed poor performance [7]. - Over the last month, the one-month average turnover rate and Fama-French three-factor one-month residual volatility factors performed well, while Beta and one-year momentum lagged [7]. - In the past year, the one-month specificity and Fama-French three-factor one-month residual volatility factors performed well, while one-year momentum and dividend yield factors underperformed [7][8]. Factor Long-Only Portfolio Performance - The long-only portfolio, constructed from the top 10% of factors, has shown cumulative excess returns, with quarterly gross margin and one-month average turnover rate factors performing well recently [9][10]. - Over the last year, small-cap stocks and earnings forecast accuracy have shown strong performance, while one-year momentum and expected adjustment averages have underperformed [9][10]. Factor Introduction - The factors used in the analysis are categorized into valuation, profitability, growth, dividends, reversal, turnover, volatility, and analyst factors, each with specific calculation methods [11][12].