Investment Rating - The industry is rated as "Neutral," indicating that the expected overall return in the next six months will be between -5% and 5% compared to the CSI 300 Index [30]. Core Insights - The report constructs a quantitative timing model based on the flow of main funds, aiming to capture short-term divergences between main fund movements and index price trends to identify potential market turning points [1][11]. - The model uses a rolling window to perform linear regression on the net inflow rate of main funds and index returns, focusing on the time-varying intercept (α) to generate buy or sell signals based on its historical extremes [2][12]. - The model has been backtested since January 2016, showing an annualized return of 12.5%, an annualized volatility of 19.8%, a Sharpe ratio of 0.63, a maximum drawdown of -39.5%, and an annualized excess return of 11.4% [3][19]. Summary by Sections Model Logic - The market's capital structure is divided into three levels: main funds, retail investors, and small investors, with significant differences in their impact on index trends. Empirical evidence shows a high positive correlation (approximately 0.84) between the net inflow rate of main funds and market index returns [11]. - At critical market turning points, there are often short-term divergences between price and main fund flows, providing potential timing opportunities [11]. Timing Framework - The timing framework is constructed based on the logic of capturing divergences between main fund flows and price trends. It includes a rolling window for linear regression to estimate the intercept (α) and compare it with historical extremes to generate trading signals [12][18]. - Specific rules are established for trading signals based on the comparison of current α with its historical extremes, with additional filters to avoid false signals during strong upward trends [16][18]. Timing Results - The current signal for the CSI 500 Index is "short," indicating a bearish outlook based on the model's analysis. The backtesting results demonstrate the model's effectiveness in identifying market trends and potential turning points [3][19].

Investment Rating - The industry is rated as "Neutral," expecting an overall return within the range of -5% to 5% compared to the CSI 300 Index over the next six months [29]. Core Insights - The report constructs a quantitative timing model based on the flow of main funds, aiming to capture short-term divergences between main fund movements and index price trends to identify potential market turning points [1][11]. - The model uses a rolling window to perform linear regression on the net inflow rate of main funds and index returns, focusing on the time series changes of the intercept α to generate buy or sell signals [2][12]. - The model has been backtested since January 2016, showing an annualized return of 8.0%, an annualized volatility of 15.8%, a Sharpe ratio of 0.51, a maximum drawdown of -35.5%, and an annualized excess return of 5.3% [3][19]. Summary by Sections Model Logic - The market's capital structure is divided into three levels: main funds, retail investors, and small investors, with significant differences in their impact on index trends. Empirical evidence shows a high positive correlation (approximately 0.84) between the net inflow rate of main funds and market index returns [11]. - Short-term divergences often occur at key market turning points, where main funds lead price stabilization at the end of declines and exit before price peaks at the end of rises, providing potential timing opportunities [11]. Timing Framework - The timing framework is built on the logic of capturing divergences between main fund flows and price trends. It involves a rolling window for linear regression to estimate the relationship between the net inflow rate and index returns [12][13]. - Specific rules are established for generating signals based on the behavior of the intercept α, with conditions to avoid false signals during strong upward trends [18] [19]. Timing Results - The current signal for the CSI 300 Index is "no position," indicating a cautious approach based on the model's findings. The backtesting results demonstrate the model's effectiveness in providing directional signals during market fluctuations [3][19].

Market Sentiment - The market sentiment indicator as of March 27 is 1.2, down from 1.7 the previous week, indicating a bearish outlook as sentiment continues to decline throughout the week [1][6]. - The price-volume consistency indicator and RSI have both turned negative this week, reflecting a shift from previous oscillation to a sustained bearish view, indicating a weakening market [1][8]. Trading Volume - The total trading volume for the A-share market decreased by 0.65% week-on-week, with an average daily trading volume of 1,394.17 billion yuan, suggesting a slight decline in market activity compared to the previous week [1][10]. Industry Performance - As of March 27, the short-term score rankings for industries show that utilities, coal, power equipment, telecommunications, and oil and petrochemicals are leading, with utilities scoring 91.53, the highest among industries, and coal scoring 84.75, second [1][34]. - The industry crowding indicator shows a low correlation of 0.17 with the weekly price changes, indicating that the crowding level is not significantly impacting price movements [1][37]. Risk Appetite - The relative trading volume of the Sci-Tech 50 index remains low, indicating that market risk appetite is also low, with a slight fluctuation observed [1][13]. - The financing balance ratio has slightly increased this week, suggesting a minor rise in market sentiment and trading activity in the financing market [1][21]. Technical Indicators - The RSI indicator has penetrated the lower boundary and continues to decline rapidly, indicating a weakening short-term momentum [1][25]. - The main buying power indicator has shown a downward trend, reflecting reduced willingness from institutional investors to actively allocate capital in the market [1][28].

Market Overview - The market is currently in a state of stock game with significant funding divergence, as indicated by the liquidity shock index for the CSI 300, which was -0.49 last Friday, lower than the previous week at 0.49, suggesting current market liquidity is above the average level of the past year by -0.49 standard deviations [1] - The PUT-CALL ratio for the SSE 50 ETF options has been rising, reaching 0.83 last Friday, up from 0.67 the previous week, indicating increased caution among investors regarding the short-term performance of the SSE 50 ETF [1] - The average turnover rates for the SSE Composite Index and Wind All A-shares were 1.38% and 1.94%, respectively, indicating a decrease in trading activity, positioned at the 78.04% and 81.46% percentile since 2005 [1] Macro Factors - The RMB exchange rate fluctuated last week, with onshore and offshore rates showing weekly declines of -0.42% and -0.2%, respectively [1] - The U.S. stock market experienced a downward trend, with the Dow Jones Industrial Average, S&P 500, and Nasdaq indices reporting weekly returns of -0.9%, -2.12%, and -3.23% [1] - The National Bureau of Statistics reported that profits of large-scale industrial enterprises in China reached 1.02 trillion yuan in January-February 2026, a year-on-year increase of 15.2%, accelerating by 14.6 percentage points compared to the previous year [1] Market Sentiment - The A-share market experienced some fluctuations and divergence last week, with heightened risk aversion due to ongoing geopolitical tensions, which have suppressed short-term risk appetite [1] - Technical analysis indicates multiple intraday reversals in the A-share market, suggesting significant funding divergence and a stock game state, leading to a low probability of upward trends in the short term [1] Factor Analysis - The overall market PE (TTM) stands at 22.5 times, positioned at the 77.5% percentile since 2005 [2] - The small-cap factor's congestion level has decreased to -0.11, while the low valuation factor's congestion level is at -0.54, indicating a shift in market dynamics [2] - Industry congestion levels are relatively high in sectors such as comprehensive, communication, non-ferrous metals, basic chemicals, and oil & petrochemicals, with the latter two sectors showing a significant increase in congestion [2]

- The report introduces a timing model based on the distance between the short-term moving average (20-day) and the long-term moving average (120-day) of the Wind All A Index. The model identifies market conditions by observing the difference between these two averages. The latest data shows the 20-day moving average at 6633 and the 120-day moving average at 6485, with a difference of 2.28%, indicating a typical consolidation phase[3][7][12] - The mid-term industry allocation model highlights sectors with strong performance trends. It suggests focusing on industries related to computing power (e.g., semiconductor equipment ETF 159516.SZ, communication ETF 515880.SH), cyclical sectors (e.g., oil and gas ETF, energy chemical ETF 159981.SH), and the new energy sector. If a volume contraction signal appears, attention should shift to non-ferrous metals and military industries[3][6][8] - The report evaluates the market's valuation levels using PE and PB metrics. The Wind All A Index PE is positioned near the 90th percentile, indicating a relatively high valuation, while the PB is at the 50th percentile, reflecting a moderate valuation level[8][12] - The timing model suggests maintaining a 50% equity allocation for absolute return products based on the Wind All A Index, considering the current market environment and valuation levels[6][8][12]

- The report constructs indicators using transaction details data to track large purchases and net active purchases[1][7] - The large order transaction amount ratio depicts the buying behavior of large funds[7] - The net active purchase amount ratio depicts investors' active buying behavior[7] - The large order transaction amount ratio is calculated by restoring transaction data to buy and sell order data and filtering large orders based on transaction volume, then calculating the ratio of large order transaction amount to the total transaction amount of the day[7] - The net active purchase amount ratio is calculated by identifying each transaction as active buy or active sell based on transaction data, subtracting the transaction amounts of the two, and calculating the ratio of net active purchase amount to the total transaction amount of the day[7] Model Backtest Results - Large order transaction amount ratio for individual stocks (20260316-20260320): Shaoneng Co., Ltd. 86.7%, Angang Steel Co., Ltd. 85.7%, Zhongli Group 85.5%, Huadian Liaohe Energy 85.5%, Wentou Holdings 85.3%, Xining Special Steel 84.9%, Jiangyan Group 84.8%, China High-Speed Railway 84.7%, Guangshen Railway 84.6%, Shaanxi International Trust 84.6%[9] - Net active purchase amount ratio for individual stocks (20260316-20260320): Yunnan Baiyao 15.5%, Supor 14.9%, ZJ Bio-Tech-U 14.5%, Industrial and Commercial Bank of China 13.9%, Fulin Precision 13.6%, China World Trade Center 13.3%, Anbotong 13.0%, Zhongwang Fabric 13.0%, Shandong Expressway 12.2%, Youngor 12.2%[10] - Large order transaction amount ratio for broad-based indices (20260316-20260320): SSE Composite Index 72.3%, SSE 50 Index 71.3%, CSI 300 Index 73.4%, CSI 500 Index 71.3%, ChiNext Index 72.4%[12] - Net active purchase amount ratio for broad-based indices (20260316-20260320): SSE Composite Index -4.6%, SSE 50 Index -4.3%, CSI 300 Index -2.3%, CSI 500 Index -3.9%, ChiNext Index 0.7%[12] - Large order transaction amount ratio for CITIC first-level industries (20260316-20260320): Petroleum and Petrochemical 76.4%, Coal 77.5%, Nonferrous Metals 73.7%, Electric Power and Public Utilities 77.5%, Steel 78.3%, Basic Chemicals 74.1%, Construction 76.9%, Building Materials 75.1%, Light Manufacturing 74.4%, Machinery 72.6%, Electric Power Equipment and New Energy 74.8%, National Defense and Military Industry 69.5%, Automotive 72.5%, Commercial Retail 74.6%, Consumer Services 74.7%, Home Appliances 75.0%, Textiles and Apparel 75.8%, Medicine 71.1%, Food and Beverage 68.7%, Agriculture, Forestry, Animal Husbandry, and Fishery 75.1%, Banking 80.0%, Non-Banking Finance 74.2%, Real Estate 77.3%, Transportation 78.3%, Electronics 69.5%, Communications 73.4%, Computers 70.5%, Media 73.3%, Comprehensive 76.1%, Comprehensive Finance 73.3%[13] - Net active purchase amount ratio for CITIC first-level industries (20260316-20260320): Petroleum and Petrochemical -3.4%, Coal 0.5%, Nonferrous Metals -4.8%, Electric Power and Public Utilities -1.0%, Steel -10.2%, Basic Chemicals -5.4%, Construction -10.0%, Building Materials -5.5%, Light Manufacturing -5.4%, Machinery -4.1%, Electric Power Equipment and New Energy -0.1%, National Defense and Military Industry -9.0%, Automotive -3.6%, Commercial Retail -12.4%, Consumer Services -4.4%, Home Appliances -5.9%, Textiles and Apparel -8.2%, Medicine -6.1%, Food and Beverage -5.1%, Agriculture, Forestry, Animal Husbandry, and Fishery -6.9%, Banking -2.2%, Non-Banking Finance -11.9%, Real Estate -8.4%, Transportation -2.3%, Electronics -2.3%, Communications 1.2%, Computers -10.9%, Media -11.4%, Comprehensive -14.2%, Comprehensive Finance -20.8%[13] - Large order transaction amount ratio for ETFs (20260316-20260320): Huatai-PineBridge CSI A500 ETF 93.6%, Huatai-PineBridge MSCI China A50 Interconnection ETF 93.5%, Guotai CSI A500 ETF 93.4%, Haifutong SSE Urban Investment Bond ETF 92.0%, Huaxia CSI A500 ETF 91.5%, Tianhong CSI Computer Theme ETF 91.2%, Guotai CSI All Index Building Materials ETF 90.4%, Southern CSI All Index Dividend Quality ETF 89.6%, Penghua CSI Oil and Natural Gas ETF 89.1%, Harvest CSI Rare Earth Industry ETF 89.1%[15] - Net active purchase amount ratio for ETFs (20260316-20260320): Tianhong CSI Industrial Nonferrous Metals Theme ETF 18.3%, Harvest CSI Green Power ETF 14.0%, Huaxia CSI Subdivided Nonferrous Metals Industry ETF 13.9%, Invesco Great Wall CSI Dividend Low Volatility 100 ETF 13.2%, Huaxia CSI Semiconductor Materials and Equipment Theme ETF 12.2%, Haifutong SSE Urban Investment Bond ETF 12.0%, Huatai-PineBridge CSI Energy ETF 11.1%, E Fund Shenzhen 100 ETF 11.0%, Southern ChiNext Artificial Intelligence ETF 10.1%, Huatai-PineBridge Dividend Low Volatility ETF 9.4%[16]

Quantitative Models and Construction Methods - **Model Name**: Convolutional Neural Network (CNN) for Price-Volume Data **Model Construction Idea**: The model leverages convolutional neural networks to analyze standardized graphical representations of price-volume data, aiming to predict future price trends. The learned features are then mapped to specific industry theme indices[76][78] **Model Construction Process**: 1. Standardize price-volume data into graphical formats for each stock within a specific time window[76] 2. Train a convolutional neural network to extract features from these graphical representations[76] 3. Map the learned features to industry theme indices, such as dividend low-volatility, banking, and real estate indices[76][78] **Model Evaluation**: The model effectively identifies industry themes based on price-volume patterns, providing actionable insights for sector allocation[76][78] Model Backtesting Results - **CNN Model**: Latest theme configurations include the following indices: 1. CSI Dividend Low Volatility Index (h30269.CSI) 2. CSI Banking Index (399986.SZ) 3. CSI 800 Banking Index (h30022.CSI) 4. CSI Mainland Real Estate Theme Index (000948.CSI) 5. CSI 800 Real Estate Index (399965.SZ)[78] Quantitative Factors and Construction Methods - **Factor Name**: Macroeconomic Indicators **Factor Construction Idea**: Macroeconomic factors are used to assess their impact on asset returns by identifying trends and significant events in historical data[51][52] **Factor Construction Process**: 1. Track 25 domestic and international macroeconomic indicators, such as PMI, CPI, PPI, and M2 growth rates[52] 2. Define four types of macroeconomic events: short-term peaks/troughs, continuous up/down trends, historical highs/lows, and trend reversals[52] 3. Use historical moving averages to classify macroeconomic trends (e.g., 3-month, 12-month averages) and analyze their impact on asset returns over the next month[54] **Factor Evaluation**: The approach identifies effective macroeconomic events that significantly influence asset returns, providing a robust framework for market trend analysis[52][54] Factor Backtesting Results - **Macroeconomic Factors**: 1. PMI (3-month moving average): Positive outlook for equities[55] 2. Social Financing Stock YoY Growth (1-month moving average): Neutral outlook[55] 3. 10-Year Treasury Yield (12-month moving average): Neutral outlook[55] 4. Dollar Index (1-month moving average): Neutral outlook[55]

- The report utilizes convolutional neural networks (CNN) to model the relationship between charted price-volume data and future prices, mapping learned features to industry thematic indices[74][75] - The thematic indices configured using CNN include the CSI Dividend Low Volatility Index, CSI Bank Index, CSI 800 Bank Index, CSI Mainland Real Estate Thematic Index, and CSI 800 Real Estate Index[75] - The CNN-based approach focuses on standardizing price-volume data into charts for analysis, as referenced in prior deep learning studies like "AI Recognition and Classification of Stock Price Trends Based on Convolutional Neural Networks"[74]

Market Sentiment Overview - As of March 20, the market sentiment indicator is at 1.7, up from 1.55 the previous week, indicating a neutral sentiment despite fluctuations throughout the week [1][4] - Multiple sub-indicators have shown a decline compared to the previous week, influenced by ongoing external political risks, suggesting a potential further drop in market sentiment [1][4] Sub-indicator Analysis - The price-volume consistency indicator has rapidly declined, reflecting a weaker correlation between price increases and market attention, indicating an overall bearish sentiment [7][9] - Total trading volume for the A-share market decreased by 12.49% week-on-week, with an average daily trading volume of 14,098.98 billion, further indicating reduced market activity [11] - The proportion of the STAR 50 index relative to the total A-share trading volume has consistently decreased, suggesting a decline in risk appetite [15] - The inter-industry trading volatility has been on the rise, reaching historical highs for 2023, indicating increased activity in switching funds between different sectors [16] - The industry trend indicator initially rose but later showed a downward trend, indicating a reduction in divergence among industry views and a slight increase in consensus on short-term value judgments [18] - The financing balance ratio has slightly decreased, indicating a reduction in market leverage and a decline in investor risk appetite [19] - The RSI indicator has penetrated the lower boundary, suggesting increased downward momentum and reduced buying power, reflecting an overall decline in market sentiment [20] - The net inflow of main funds has shown a downward trend, indicating weakened buying power and reduced enthusiasm from institutional investors [24] Industry Crowding and Trading Heat - The highest average crowding levels as of March 20 are in the utilities, basic chemicals, electrical equipment, construction decoration, and environmental protection sectors, while the lowest are in automotive, defense, social services, retail, and textiles [30][31] - The correlation between crowding and weekly price changes is near zero, indicating that high crowding does not necessarily lead to price increases, with sectors like construction decoration and environmental protection showing low price changes despite high crowding [32] Trend Scoring Model Insights - The short-term scoring model indicates that sectors such as coal, utilities, electrical equipment, communication, and construction decoration are leading in trend scores, with coal having the highest score of 93.22 [25][28] - The model suggests a preference for growth and large-cap styles, with the current signals indicating a strong preference for large-cap stocks [35]

Core Viewpoint - The A-share market is expected to remain in a state of fluctuation in the short term, as indicated by various technical and quantitative indicators [1][2]. Market Overview - During the week of March 16-20, 2026, the Shanghai Composite Index fell by 2.47%, the CSI 300 Index decreased by 2.19%, the CSI 500 Index dropped by 5.82%, while the ChiNext Index rose by 1.26% [3]. - The current overall market PE (TTM) stands at 22.6 times, which is at the 78.3% percentile since 2005 [3]. - Historical data shows that the CSI 500 Index has performed well in the latter half of March since 2005 [3]. Factor Crowding Observation - The small-cap factor crowding has increased, with a current value of 0.09. The low valuation factor crowding is at -0.31, while the high profitability factor crowding is at 0.24, and the high profitability growth factor crowding is at 0.25 [3]. Industry Crowding - Industries such as comprehensive, communication, non-ferrous metals, steel, and electronics exhibit relatively high crowding levels. The oil and petrochemical, as well as agriculture, forestry, animal husbandry, and fishery industries have seen a significant increase in crowding [4].