Quantitative Models and Construction Methods - **Model Name**: Industry Allocation Model **Model Construction Idea**: This model aims to recommend industry sectors based on medium-term perspectives, focusing on sectors with potential for recovery or growth trends[2][3][10] **Model Construction Process**: The model identifies sectors with recovery potential ("困境反转型板块") and growth opportunities. It recommends sectors such as healthcare (恒生医疗), export-related consumer sectors (e.g., light industry and home appliances), and technology sectors (信创, communication, solid-state batteries). Additionally, it highlights sectors with ongoing upward trends, such as banking and gold[2][3][10] **Model Evaluation**: The model provides actionable insights for medium-term industry allocation, emphasizing sectors with recovery potential and growth trends[2][3][10] - **Model Name**: TWO BETA Model **Model Construction Idea**: This model focuses on identifying technology-related sectors with growth potential[2][3][10] **Model Construction Process**: The TWO BETA model recommends technology sectors, including 信创, communication, and solid-state batteries, based on their growth potential and market trends[2][3][10] **Model Evaluation**: The model effectively identifies technology sectors with strong growth potential, aligning with market trends[2][3][10] - **Model Name**: Timing System Model **Model Construction Idea**: This model evaluates market conditions by analyzing the distance between short-term and long-term moving averages to determine market trends[2][9][14] **Model Construction Process**: 1. Define the short-term moving average (20-day) and long-term moving average (120-day) for the Wind All A Index 2. Calculate the difference between the two moving averages: $ \text{Difference} = \text{20-day MA} - \text{120-day MA} $ - Latest values: 20-day MA = 4946, 120-day MA = 5088 - Difference = -2.80% (previous week: -3.63%) 3. Monitor the absolute value of the difference; when it falls below 3%, the market is considered to be in a consolidation phase[2][9][14] **Model Evaluation**: The model provides a clear signal for market consolidation, aiding in timing decisions[2][9][14] - **Model Name**: Position Management Model **Model Construction Idea**: This model determines the recommended equity allocation based on valuation levels and short-term market trends[3][10] **Model Construction Process**: 1. Assess valuation levels of the Wind All A Index: - PE ratio: 50th percentile (medium level) - PB ratio: 10th percentile (low level) 2. Combine valuation levels with short-term market trends to recommend a 60% equity allocation for absolute return products[3][10] **Model Evaluation**: The model provides a systematic approach to position management, balancing valuation and market trends[3][10] Backtesting Results of Models - **Industry Allocation Model**: No specific numerical backtesting results provided[2][3][10] - **TWO BETA Model**: No specific numerical backtesting results provided[2][3][10] - **Timing System Model**: - Latest moving average difference: -2.80% - Previous week difference: -3.63% - Absolute difference < 3%, indicating a consolidation phase[2][9][14] - **Position Management Model**: - Recommended equity allocation: 60%[3][10]

Quantitative Models and Construction Methods 1. Model Name: Market Sentiment Timing Model - **Model Construction Idea**: This model is designed to quantify market sentiment using a structured approach, incorporating multiple sub-indicators to assess the overall sentiment direction [7][8] - **Model Construction Process**: 1. Sub-indicators used include: industry trading volatility, industry trading congestion, price-volume consistency, Sci-Tech 50 trading proportion, industry trend, RSI, main buying force, PCR combined with VIX, and financing balance proportion [8] 2. Each sub-indicator is scored based on its sentiment direction and position within Bollinger Bands, with scores categorized as (-1, 0, 1) [8] 3. The final sentiment structure indicator is calculated as the 20-day moving average of the summed scores, oscillating around the zero axis within the range of [-6, 6] [8] - Formula: $ \text{Sentiment Indicator} = \text{20-day MA of (Sum of Sub-indicator Scores)} $ - **Model Evaluation**: The model effectively captures market sentiment fluctuations, with significant sentiment recovery observed since April 2024 [8][9] 2. Model Name: Moving Average Sequence Scoring (MASS) Model - **Model Construction Idea**: This model evaluates the long-term and short-term trends of indices by analyzing the relative positions of moving averages over different time horizons [20] - **Model Construction Process**: 1. For a given period \( N \) (e.g., \( N=360 \) for long-term, \( N=60 \) for short-term), calculate scores for \( N \) moving averages [20] 2. If a shorter moving average \( k \) is above the longer moving average \( k+1 \), assign a score of 1; otherwise, assign 0 [20] 3. Normalize the scores to a range of 0-100 and compute the average score for the index at a specific time point [20] 4. Calculate the 100-day and 20-day moving averages of the trend scores to generate buy/sell signals [20] - Formula: $ \text{Trend Score} = \frac{\text{Sum of Scores}}{N} \times 100 $ - **Model Evaluation**: The model provides clear signals for trend reversals, with recent results indicating a shift towards growth-oriented sectors [20][21] 3. Model Name: RSI Style Timing Model - **Model Construction Idea**: This model uses the Relative Strength Index (RSI) to evaluate the relative strength of different market styles (e.g., growth vs. value, small-cap vs. large-cap) [24] - **Model Construction Process**: 1. Calculate the net value ratio of two style indices (e.g., growth/value) over a fixed period [24] 2. Compute the RSI using the formula: $ \text{RSI} = 100 - \frac{100}{1 + \frac{\text{Average Gain}}{\text{Average Loss}}} $ - Where "Gain" represents average positive changes, and "Loss" represents average negative changes over \( N \) days [24] 3. Compare the 20-day RSI with the 60-day RSI to determine the dominant style [24] - **Model Evaluation**: The model indicates a clear shift from large-cap value to small-cap growth styles, with strong confirmation from recent RSI trends [24][27] --- Model Backtesting Results 1. Market Sentiment Timing Model - Sentiment Indicator Value: 1.5 as of May 9, 2025, indicating a positive sentiment recovery [9] 2. Moving Average Sequence Scoring (MASS) Model - Short-term signals: Positive for indices such as CSI 300, CSI A500, and ChiNext, with short-term scores ranging from 33.90 to 40.68 [36] - Long-term signals: Positive for most indices, with long-term scores exceeding 66.57 for indices like ChiNext [36] 3. RSI Style Timing Model - Growth/Value RSI: Growth-dominant with RSI values of 57.91 (short-term) and 55.24 (long-term) for the CSI Growth/Value index [27] - Small/Large Cap RSI: Small-cap dominant with RSI values of 59.84 (short-term) and 60.16 (long-term) for the Small/Large Cap index [27] --- Quantitative Factors and Construction Methods 1. Factor Name: Price-Volume Consistency - **Factor Construction Idea**: Measures the stability of market sentiment based on the alignment of price and volume movements [8] - **Factor Construction Process**: 1. Calculate the correlation between price changes and trading volume over a fixed period [8] 2. Assign scores based on the strength of the correlation, with higher scores indicating stronger consistency [8] - **Factor Evaluation**: The factor showed significant improvement in recent weeks, contributing to the overall sentiment recovery [11][16] 2. Factor Name: RSI - **Factor Construction Idea**: Reflects the relative strength of buying vs. selling pressure over a specific period [24] - **Factor Construction Process**: 1. Compute average gains and losses over \( N \) days [24] 2. Use the RSI formula to calculate the index value [24] - **Factor Evaluation**: RSI values above 50 indicate strong buying pressure, with recent results favoring growth and small-cap styles [24][27] --- Factor Backtesting Results 1. Price-Volume Consistency - Recent Score: Increased to 1 as of May 9, 2025, indicating improved alignment between price and volume [12] 2. RSI - Growth/Value RSI: Growth-dominant with short-term RSI of 57.91 [27] - Small/Large Cap RSI: Small-cap dominant with short-term RSI of 59.84 [27]

Market Performance - The recent 5 trading days saw the Sci-Tech 50 Index increase by 0.24%, the ChiNext Index rise by 4.13%, large-cap value stocks up by 1.55%, large-cap growth stocks up by 2.05%, the SSE 50 Index up by 1.46%, and the small-cap represented by the CSI 2000 up by 3.77% [1] - The defense and military industry, as well as the communication sector, performed well, while steel and retail sectors lagged behind [1] Risk Premium Analysis - The static PE of the CSI All Index minus the yield of 10-year government bonds indicates a risk premium, which has historically reached extreme levels at two standard deviations above the mean during significant market bottoms, such as in 2012, 2018, and 2020 [1] - As of April 26, 2022, the risk premium reached 4.17%, and on October 28, 2022, it was 4.08%, with a recent reading of 4.11% on January 19, 2024, marking the fifth occurrence since 2016 of exceeding 4% [1] Valuation Levels - As of May 9, 2025, the CSI All Index's PETTM is at the 50th percentile, with the SSE 50 and CSI 300 at 61% and 47% respectively, while the ChiNext Index is close to 11% [2] - The ChiNext Index's valuation is relatively low compared to historical averages [2] Long-term Market Trends - The technical analysis of the Deep 100 Index indicates a pattern of bear markets every three years followed by bull markets, with previous declines ranging from 40% to 45% [2] - The current adjustment cycle began in Q1 2021, suggesting a potential for upward movement from the bottom [2] Fund Flow and Trading Activity - In the last 5 trading days, ETF funds saw an outflow of 17.9 billion yuan, while margin trading increased by approximately 4.4 billion yuan [2] - The average daily trading volume across both markets was 1.2918 trillion yuan [2] AI and Machine Learning Insights - A convolutional neural network (CNN) was utilized to model price and volume data, mapping learned features to industry themes, with a current focus on banking [2][7] Market Sentiment - The proportion of stocks above the 200-day moving average is being tracked to gauge market sentiment [9] Equity and Bond Risk Preference - Ongoing monitoring of risk preferences between equity and bond assets is being conducted [11]

Investment Rating - The industry investment rating is "Neutral" with an expected industry index increase of -5% to 5% relative to the CSI 300 index over the next six months [22]. Core Insights - The market is currently in a downtrend, with a focus on when the profit effect will turn positive. The current profit effect is around -1% [2][10]. - The report suggests maintaining a neutral position until the 30-day moving average of the wind All A index is breached, considering the low valuation levels [4][10]. - The industry configuration model recommends focusing on "dilemma reversal" sectors, particularly in healthcare and consumer sectors related to export chains such as light industry and home appliances [3][10]. - The TWO BETA model continues to recommend the technology sector, emphasizing domestic substitution in the fields of information technology and AI chips [3][10]. - Despite a significant drop on Friday, the banking sector, which is still in an upward trend, remains worthy of attention [3][10]. Summary by Sections Market Overview - The wind All A index is currently in a downtrend, with the 20-day moving average at 4908 and the 120-day moving average at 5092.8, indicating a distance of -3.63% [2][9]. - The market's current environment is characterized by uncertainty due to upcoming Federal Reserve meetings and the release of April import and export data [4][10]. Valuation Metrics - The overall PE ratio of the wind All A index is around the 50th percentile, indicating a medium level, while the PB ratio is around the 20th percentile, indicating a relatively low level [3][10]. Positioning Recommendations - The report advises a 50% allocation in absolute return products based on the wind All A index as the main stock allocation [3][10].

Quantitative Models and Construction Methods 1. Model Name: Market Sentiment Timing Model - **Model Construction Idea**: The model is built from a structural perspective to quantify market sentiment using various sub-indicators[7] - **Model Construction Process**: - The model uses sub-indicators such as industry trading volatility, trading crowding, price-volume consistency, Sci-Tech Innovation Board (STAR 50) trading proportion, industry trend, RSI, main buying force, PCR combined with VIX, and financing balance ratio[8] - Each sub-indicator is scored based on its sentiment direction and position within Bollinger Bands. Scores are categorized as (-1, 0, 1)[8] - The final sentiment structural indicator is the 20-day moving average of the summed scores. The indicator fluctuates around 0 within the range of [-6, 6][8] - **Model Evaluation**: The model effectively captures market sentiment trends and provides actionable insights for timing decisions[8] 2. Model Name: Moving Average Scoring System (MASS) - **Model Construction Idea**: This model evaluates long-term and short-term trends of indices using N-day moving averages to generate timing signals[18] - **Model Construction Process**: - For N moving averages (N=360 for long-term, N=60 for short-term), scores are assigned based on the relative position of adjacent moving averages. If a shorter moving average is above a longer one, it scores 1; otherwise, it scores 0[18] - The scores are standardized to a 0-100 scale and averaged to derive the trend score at a specific time point[18] - Long/short-term timing signals are generated based on the crossover of the trend score with its 100/20-day moving average[18] - **Model Evaluation**: The model provides clear signals for sector rotation and market style preferences, favoring value and defensive sectors in the current environment[18] 3. Model Name: RSI Style Timing Model - **Model Construction Idea**: The model uses the Relative Strength Index (RSI) to compare the relative strength of different market styles (e.g., growth vs. value, small-cap vs. large-cap)[22] - **Model Construction Process**: - For two indices A and B, calculate the standardized ratio of their net values over a fixed period[22] - Compute the average gain (Gain) and average loss (Loss) over N days, where gains on down days are treated as 0 and losses on up days are treated as 0[22] - RSI formula: $ RSI = 100 - 100 / (1 + Gain / Loss) $ - RSI values range from 0 to 100, with values above 50 indicating stronger buying pressure[22] - The model calculates 5-day, 20-day, and 60-day RSI values. When the 20-day RSI exceeds the 60-day RSI, the numerator style is favored; otherwise, the denominator style is favored[22] - **Model Evaluation**: The model effectively identifies style dominance, currently favoring large-cap and value styles while noting short-term strengthening of growth and small-cap styles[22] --- Model Backtesting Results 1. Market Sentiment Timing Model - Sentiment indicator value as of April 30, 2025: 0.8, indicating a recovery in market sentiment[9] 2. Moving Average Scoring System (MASS) - Short-term signals: Positive for sectors like beauty care (72.88), utilities (86.44), banking (74.58), and oil & petrochemicals (22.03)[19] - Long-term signals: Positive for sectors like banking (95.54), machinery (78.55), and steel (51.25)[19] 3. RSI Style Timing Model - Growth/Value (300 Growth/300 Value): RSI 20-day = 53.02, RSI 60-day = 50.42, favoring value[25] - Small-cap/Large-cap (SW Small/SW Large): RSI 20-day = 48.84, RSI 60-day = 53.62, favoring large-cap[25] --- Quantitative Factors and Construction Methods 1. Factor Name: RSI - **Factor Construction Idea**: Measures the relative strength of buying and selling forces over a specific period[22] - **Factor Construction Process**: - Calculate the average gain (Gain) and average loss (Loss) over N days[22] - Formula: $ RSI = 100 - 100 / (1 + Gain / Loss) $ - RSI values range from 0 to 100, with higher values indicating stronger buying pressure[22] - **Factor Evaluation**: Provides a robust measure of market momentum and style preferences[22] --- Factor Backtesting Results 1. RSI - Growth/Value (300 Growth/300 Value): RSI 20-day = 53.02, RSI 60-day = 50.42, favoring value[25] - Small-cap/Large-cap (SW Small/SW Large): RSI 20-day = 48.84, RSI 60-day = 53.62, favoring large-cap[25]

金融工程 | 金工定期报告 2025 年 05 月 05 日 作者 吴先兴 分析师 SAC 执业证书编号：S1110516120001 wuxianxing@tfzq.com 相关报告 1 《金融工程：金融工程-因子跟踪周 报 ： Beta 、换手率因子表现较好 -20250504》 2025-05-04 2 《金融工程：金融工程-哪些行业进 入高估区域？——估值与基金重仓股配 置监控 2025-05-03》 2025-05-03 3 《金融工程：金融工程-净利润断层 本周超额基准 0.92%》 2025-05-03 金融工程 证券研究报告 量化择时周报：突破压力位前保持中性 突破压力位前保持中性 上周周报(20250427)认为：全 A 指数的 30 日均线构成压力位，但考虑到估 值不高，建议在压力位突破前维持中性仓位。最终 wind 全 A 维持原状。 市值维度上，上周代表小市值股票的中证 2000 上涨 0.84%，中盘股中证 500 上涨 0.08%，沪深 300 下跌 0.43%，上证 50 下跌 0.59%;上周中信一级行业中， 表现较强行业包括传媒、计算机，传媒上涨 2.86%，综合金融、房地产 ...

市场情绪自3月20日持续调整，于4月18日下降至低点，数值为0.1。本周市场情绪指标在接近0轴处开始向上反弹，回升至0.5，数值较上周五(4/18)上升0.4，模型转多，市场 情绪有所缓和。 本周A股市场提示市场情绪有一定修复，较上周明显发生变化的指标有科创50成交占比、主力买入力量和期权波动率。主力流出速率减缓和VIX指标体现的恐慌程度减弱是本 周市场情绪回升的主要原因。 科创50成交占比、行业涨跌趋势性、主力买入力量和PCR结合VIX，分别代表了市场风险偏好程度下降，市场情绪不确定性增强，主力流出速度 减缓和期权市场恐慌情绪缓和。其他指标维持和上周一致的判断。 资金当前对成长高估值板块观点不确定性较强。 自上周科创50成交占比指标快速下跌至下轨以下后，本周科创50成交占比指标仍在持续下降。本周主力资金持续从科创板块 流出，累计净流出超过32亿人民币。 投资者信心逐渐恢复，市场的活跃度和投资者参与度都有了明显提升。 除了看到主力资金本周流出科创板，主力资金本周在全A仍然呈现净流出的态势，但流出速度较上周有 所减缓，主力流出主力买入力量指标有所回升。从主力资金净流出绝对量看，主力资金本周累计净流出超过370亿 ...

1. 情绪模型观点：市场情绪进一步下行 根 据 《 从 结 构 化 视 角 全 新 打 造 市 场 情 绪 择 时 模 型 》 文 中 提 到 的 构 建 思 路 ， 目 前 我 们 用 于 构 建 市 场 情 绪 结 构 指 标 所 用 到 的 细 分 指 标 如 下 表 | 指标简称 | 含义 | 情绪指示方向 | | --- | --- | --- | | 行业间交易波动率 | 资金在各板块间的交易活跃度 | 正向 | | 行业交易拥挤度 | 极值状态判断市场是否过热 | 负向 | | 价量一致性 | 资金情绪稳定性 | 正向 | | 科创 50 成交占比 | 资金风险偏好 | 正向 | | 行业涨跌趋势性 | 刻画市场轮涨补涨程度，趋势衡量 | 正向 | | RSI | 价格体现买方和卖方力量相对强弱 | 正向 | | 主力买入力量 | 主力资金净流入水平 | 正向 | | PCR 结合 VIX | 从期权指标看市场多空情绪 | 正向或负向 | | 融资余额占比 | 资金对当前和未来观点多空 | 6 公众号 · 普罗完酒会工 | 在指标合成方法上，模型采用打分的方式，根据每个分项指标所提示的情绪方向和 ...

1. 情绪模型观点：市场情绪进一步下行 根据《从结构化视角全新打造市场情绪择时模型》文中提到的构建思路，目前我们用于构建市场情绪结构指标所用到 的细分指标如下表 在指标合成方法上，模型采用打分的方式，根据每个分项指标所提示的情绪方向和所处布林轨道位置计算各指标分 数，指标分数可分为（-1，0，1）三种情况，最终对各个指标分数等权求和。最终的情绪结构指标为求和后分数的20 日均线，如图1所示，指标整体围绕0轴在[-6,6]的范围内上下波动，近5年A股市场情绪波动较大，其中2023年大部分 时间指标都处于较低位置，直至2024年10月市场情绪得分突破2。 市场情绪自3月20日持续调整，当前已下降接近0轴，为0.1，数值较上周五(4/11)下降0.4，模型维持看空观点。 1.1 从分项指标出发：市场进一步缩量，资金不确定情绪增长 本周A股市场继续提示市场情绪下行，速度没有呈现减缓趋势。本周市场情绪不确定性增强，风险偏好程度下降是市 场情绪进一步调整的主要原因。 下表展示了4月以来的情绪结构各分项指标的分数情况，从分项指标出发，本周明显提示信号切换的指标为科创50成 交占比和300RSI指标，分别代表了市场风险偏好程 ...

- Model Name: Short-term Quantitative Timing Model; Model Construction Idea: The model uses various market indicators to generate signals for market timing; Model Construction Process: The model integrates valuation, liquidity, fundamental, and sentiment signals to determine market timing. For example, the sentiment signal is derived from the volume sentiment indicator, which is constructed using the 60-day Bollinger Bands of trading volume and turnover rate. The formula for the volume sentiment score is a linear mapping of the 60-day average within the range of -1 to +1, with extreme values capped at -1 or +1. The weekly average of the 5-year percentile is used as one of the timing judgment signals. If the percentile is greater than 60%, it indicates strong sentiment and gives an optimistic signal; if less than 40%, it indicates weak sentiment and gives a cautious signal; if between 40%-60%, it gives a neutral signal. The formula is: $$ \text{Volume Sentiment Score} = \frac{\text{Current Value} - \text{Mean}}{\text{Standard Deviation}} $$ where the mean and standard deviation are calculated over a 60-day period[21][22][23]; Model Evaluation: The model has shown predictive power for the market's performance in the following week[21][22][23] - Model Name: Growth-Value Style Rotation Model; Model Construction Idea: The model suggests overweighting growth or value styles based on economic cycle analysis; Model Construction Process: The model uses the slope of the profit cycle, the level of the interest rate cycle, and the trend of the credit cycle to determine the style allocation. For example, a steep profit cycle slope and low interest rate cycle level favor growth, while a weakening credit cycle favors value. The model also considers valuation differences, such as the 5-year percentile of the PE and PB valuation differences between growth and value. The formula for the PE valuation difference is: $$ \text{PE Valuation Difference} = \frac{\text{PE of Growth} - \text{PE of Value}}{\text{PE of Value}} $$ The model gives signals based on these indicators, suggesting overweighting growth if the indicators favor growth and vice versa[39][40][41]; Model Evaluation: The model has significantly outperformed the benchmark since the end of 2012, with an annualized return of 11.44% compared to the benchmark's 6.59%[40][43] - Model Name: Small-Cap vs. Large-Cap Style Rotation Model; Model Construction Idea: The model suggests balanced allocation between small-cap and large-cap styles based on economic cycle analysis; Model Construction Process: The model uses the slope of the profit cycle, the level of the interest rate cycle, and the trend of the credit cycle to determine the style allocation. For example, a steep profit cycle slope and low interest rate cycle level favor small-cap, while a weakening credit cycle favors large-cap. The model also considers valuation differences, such as the 5-year percentile of the PE and PB valuation differences between small-cap and large-cap. The formula for the PB valuation difference is: $$ \text{PB Valuation Difference} = \frac{\text{PB of Small-Cap} - \text{PB of Large-Cap}}{\text{PB of Large-Cap}} $$ The model gives signals based on these indicators, suggesting balanced allocation if the indicators favor both styles equally[44][45][46]; Model Evaluation: The model has significantly outperformed the benchmark since the end of 2012, with an annualized return of 12.32% compared to the benchmark's 6.74%[45][47] - Model Name: Four-Style Rotation Model; Model Construction Idea: The model combines the conclusions of the growth-value and small-cap vs. large-cap rotation models to recommend allocation among four styles; Model Construction Process: The model integrates the signals from the growth-value and small-cap vs. large-cap models to determine the allocation among small-cap growth, small-cap value, large-cap growth, and large-cap value. The recommended allocation is based on the latest signals from the individual models. For example, if both models favor growth and small-cap, the allocation would be higher for small-cap growth. The formula for the combined allocation is: $$ \text{Allocation} = \text{Weight from Growth-Value Model} \times \text{Weight from Small-Cap vs. Large-Cap Model} $$ The model gives signals based on these combined indicators[48][49][50]; Model Evaluation: The model has significantly outperformed the benchmark since the end of 2012, with an annualized return of 13.10% compared to the benchmark's 7.15%[48][49][50] Model Backtest Results - Short-term Quantitative Timing Model: Annualized Return 16.39%, Annualized Volatility 14.75%, Maximum Drawdown 27.70%, Sharpe Ratio 0.9675, IR 0.5918[28][32][35] - Growth-Value Style Rotation Model: Annualized Return 11.44%, Annualized Volatility 20.87%, Maximum Drawdown 43.07%, Sharpe Ratio 0.5285, IR 0.2657[40][43] - Small-Cap vs. Large-Cap Style Rotation Model: Annualized Return 12.32%, Annualized Volatility 22.72%, Maximum Drawdown 50.65%, Sharpe Ratio 0.5377, IR 0.2432[45][47] - Four-Style Rotation Model: Annualized Return 13.10%, Annualized Volatility 21.59%, Maximum Drawdown 47.91%, Sharpe Ratio 0.5864, IR 0.2735[48][49][50]