- The report highlights the strong performance of high-frequency factors, with notable multi-granularity factor returns and differentiated excess returns in AI-enhanced portfolios [6][10][11] - High-frequency factors such as intraday skewness, downside volatility proportion, and opening buy intention proportion recorded significant multi-long-short returns and excess returns across weekly, monthly, and YTD periods [6][10][11] - Multi-granularity models (5-day and 10-day labels) achieved robust multi-long-short returns, with the 5-day label model yielding 0.91% (weekly), 2.6% (monthly), and 10.22% (YTD), while the 10-day label model delivered 0.71% (weekly), 3.06% (monthly), and 8.9% (YTD) [6][10][11] - AI-enhanced portfolios, including the Air Index Increment and CSI 500/1000 AI Enhanced portfolios, demonstrated varying excess and absolute returns under both wide and strict constraints, with weekly and daily rebalancing strategies [6][15][73] - The optimization objective for AI-enhanced portfolios is to maximize expected returns, represented by the function: $$max\sum\mu_{i}w_{i}$$ where \(w_i\) is the weight of stock \(i\) in the portfolio, and \(\mu_i\) is the expected excess return of stock \(i\) [75][76]

Quantitative Models and Construction Methods - **Model Name**: CSI 300 Enhanced Portfolio **Model Construction Idea**: Enhance the performance of the CSI 300 index by leveraging quantitative strategies[5][13] **Model Construction Process**: The model constructs an enhanced portfolio based on quantitative factors and optimization techniques. It aims to generate excess returns over the CSI 300 index by selecting stocks with favorable factor exposures and minimizing tracking error relative to the benchmark[13][15] **Model Evaluation**: The model demonstrated positive excess returns over the benchmark in the short term and year-to-date performance, indicating effective factor selection and portfolio construction[15] - **Model Name**: CSI 500 Enhanced Portfolio **Model Construction Idea**: Enhance the performance of the CSI 500 index using quantitative factor-based strategies[5][13] **Model Construction Process**: Similar to the CSI 300 Enhanced Portfolio, this model selects stocks with favorable factor exposures and optimizes the portfolio to achieve excess returns while maintaining low tracking error relative to the CSI 500 index[13][15] **Model Evaluation**: The model showed mixed results, with short-term underperformance but positive year-to-date excess returns, suggesting room for improvement in factor selection or portfolio optimization[15] - **Model Name**: CSI 1000 Enhanced Portfolio **Model Construction Idea**: Enhance the performance of the CSI 1000 index by applying quantitative factor-based strategies[5][13] **Model Construction Process**: The model employs a similar approach to the CSI 300 and CSI 500 Enhanced Portfolios, focusing on factor-based stock selection and portfolio optimization to achieve excess returns over the CSI 1000 index[13][15] **Model Evaluation**: The model achieved positive excess returns both in the short term and year-to-date, indicating effective implementation of the strategy[15] Model Backtesting Results - **CSI 300 Enhanced Portfolio**: - Weekly return: -0.66% - Weekly excess return: 0.75% - Year-to-date return: 2.53% - Year-to-date excess return: 5.28%[5][13][15] - **CSI 500 Enhanced Portfolio**: - Weekly return: -0.47% - Weekly excess return: -0.18% - Year-to-date return: 3.06% - Year-to-date excess return: -0.58%[5][13][15] - **CSI 1000 Enhanced Portfolio**: - Weekly return: -0.25% - Weekly excess return: 0.23% - Year-to-date return: 4.77% - Year-to-date excess return: 2.78%[5][13][15] Quantitative Factors and Construction Methods - **Factor Name**: Market Capitalization (Size Factor) **Factor Construction Idea**: Capture the performance difference between small-cap and large-cap stocks[47] **Factor Construction Process**: Stocks are sorted by market capitalization, and the top 10% (small-cap) and bottom 10% (large-cap) are selected to form long and short portfolios, respectively. The difference in returns between these portfolios represents the factor's performance[46][47] **Factor Evaluation**: Small-cap stocks outperformed large-cap stocks, with positive multi-period excess returns, indicating the effectiveness of the size factor[47][48] - **Factor Name**: PB (Price-to-Book Ratio) **Factor Construction Idea**: Identify undervalued stocks by comparing their market price to book value[47] **Factor Construction Process**: Stocks are ranked by PB ratio, and the top 10% (low PB) and bottom 10% (high PB) are selected to form long and short portfolios, respectively. The difference in returns between these portfolios represents the factor's performance[46][47] **Factor Evaluation**: The PB factor showed mixed results, with positive returns in some periods and underperformance in others, suggesting sensitivity to market conditions[47][48] - **Factor Name**: Turnover Rate **Factor Construction Idea**: Measure investor activity and sentiment by analyzing stock turnover rates[51] **Factor Construction Process**: Stocks are ranked by turnover rate, and the top 10% (low turnover) and bottom 10% (high turnover) are selected to form long and short portfolios, respectively. The difference in returns between these portfolios represents the factor's performance[46][51] **Factor Evaluation**: The turnover rate factor contributed positive returns, particularly in the short term, indicating its relevance in capturing market sentiment[51][54] - **Factor Name**: Expected Net Profit Adjustment **Factor Construction Idea**: Reflect the market's expectations of future profitability adjustments[56] **Factor Construction Process**: Stocks are ranked by expected net profit adjustments, and the top 10% (high adjustment) and bottom 10% (low adjustment) are selected to form long and short portfolios, respectively. The difference in returns between these portfolios represents the factor's performance[46][56] **Factor Evaluation**: The factor consistently delivered positive returns, highlighting its effectiveness in capturing market expectations[56][57] Factor Backtesting Results - **Market Capitalization (Size Factor)**: - Weekly long-short return: 2.51% - Year-to-date long-short return: 6.09%[47][48] - **PB Factor**: - Weekly long-short return: -0.62% - Year-to-date long-short return: 3.74%[47][48] - **Turnover Rate Factor**: - Weekly long-short return: 0.27% - Year-to-date long-short return: 6.09%[51][54] - **Expected Net Profit Adjustment Factor**: - Weekly long-short return: 0.50% - Year-to-date long-short return: 2.45%[56][57]

Quantitative Models and Construction Methods 1. Model Name: Multi-Granularity Model (5-Day Label) - **Model Construction Idea**: This model leverages deep learning techniques to capture multi-granularity features of stock data over a 5-day horizon[66] - **Model Construction Process**: The factor is trained using a bidirectional AGRU (Attention Gated Recurrent Unit) model, which processes sequential data to extract temporal dependencies and patterns[66] - **Model Evaluation**: The model demonstrates stable performance across different time periods, indicating its robustness in capturing market dynamics[66] 2. Model Name: Multi-Granularity Model (10-Day Label) - **Model Construction Idea**: Similar to the 5-day label model, this model extends the horizon to 10 days to capture longer-term patterns in stock data[70] - **Model Construction Process**: The factor is also trained using a bidirectional AGRU model, with adjustments to accommodate the extended time horizon[70] - **Model Evaluation**: The model shows consistent performance, with slightly different characteristics compared to the 5-day label model, making it suitable for longer-term strategies[70] 3. Model Name: AI-Enhanced Index Strategies - **Model Construction Idea**: Combines multiple deep learning factors (e.g., 5-day and 10-day multi-granularity models) to construct AI-enhanced index strategies with risk constraints[72] - **Model Construction Process**: - The combined factor is a weighted sum: `0.5 * Multi-Granularity Model (5-Day Label) + 0.5 * Multi-Granularity Model (10-Day Label)`[72] - Optimization objective: Maximize expected returns, represented by the function: $$ max \sum \mu_{i} w_{i} $$ where \( w_{i} \) is the weight of stock \( i \), and \( \mu_{i} \) is the expected excess return of stock \( i \)[75] - Risk control constraints include limits on individual stock weights, industry weights, market capitalization, and turnover rates[73][75] - Backtesting assumes next-day average price execution and deducts a 0.3% transaction cost[76] - **Model Evaluation**: The model effectively balances return maximization and risk control, with different configurations (e.g., wide vs. strict constraints) tailored to specific index benchmarks[72][73] --- Model Backtesting Results 1. Multi-Granularity Model (5-Day Label) - **IC**: Historical: 0.079; 2026: 0.040[14] - **e^(-RankMAE)**: Historical: 0.343; 2026: 0.334[14] - **Long-Short Return**: March: 1.68%; 2026 YTD: 9.31%[14] - **Long-Only Excess Return**: March: 1.21%; 2026 YTD: 4.95%[14] - **Monthly Win Rate**: 9/10[14] 2. Multi-Granularity Model (10-Day Label) - **IC**: Historical: 0.072; 2026: 0.040[14] - **e^(-RankMAE)**: Historical: 0.342; 2026: 0.336[14] - **Long-Short Return**: March: 2.35%; 2026 YTD: 8.19%[14] - **Long-Only Excess Return**: March: 1.48%; 2026 YTD: 4.72%[14] - **Monthly Win Rate**: 8/10[14] 3. AI-Enhanced Index Strategies - **AI Air Quality Index Strategy**: - **Weekly Rebalancing**: Excess Return: -0.12% (last week), 0.65% (March), 4.17% (2026 YTD); Absolute Return: -5.47% (last week), -7.86% (March), 6.70% (2026 YTD)[15][81] - **Daily Rebalancing**: Excess Return: -0.78% (last week), -0.08% (March), 4.41% (2026 YTD); Absolute Return: -6.12% (last week), -8.59% (March), 6.94% (2026 YTD)[15][81] - **CSI 500 AI Enhanced (Wide Constraint)**: - **Weekly Rebalancing**: Excess Return: 1.43% (last week), 5.62% (March), 2.71% (2026 YTD); Absolute Return: -4.40% (last week), -4.76% (March), 6.66% (2026 YTD)[15][83] - **Daily Rebalancing**: Excess Return: 0.60% (last week), 1.79% (March), -2.71% (2026 YTD); Absolute Return: -5.23% (last week), -8.58% (March), 1.24% (2026 YTD)[15][83] - **CSI 500 AI Enhanced (Strict Constraint)**: - **Weekly Rebalancing**: Excess Return: 0.35% (last week), 3.51% (March), 2.73% (2026 YTD); Absolute Return: -5.47% (last week), -6.87% (March), 6.68% (2026 YTD)[15][89] - **Daily Rebalancing**: Excess Return: 0.31% (last week), 2.10% (March), 1.42% (2026 YTD); Absolute Return: -5.52% (last week), -8.27% (March), 5.37% (2026 YTD)[15][89] - **CSI 1000 AI Enhanced (Wide Constraint)**: - **Weekly Rebalancing**: Excess Return: 0.79% (last week), 3.52% (March), 4.19% (2026 YTD); Absolute Return: -4.46% (last week), -5.56% (March), 6.67% (2026 YTD)[15][91] - **Daily Rebalancing**: Excess Return: -0.20% (last week), 1.81% (March), 1.92% (2026 YTD); Absolute Return: -5.44% (last week), -7.27% (March), 4.40% (2026 YTD)[15][91] - **CSI 1000 AI Enhanced (Strict Constraint)**: - **Weekly Rebalancing**: Excess Return: 0.57% (last week), 2.55% (March), 3.67% (2026 YTD); Absolute Return: -4.68% (last week), -6.53% (March), 6.15% (2026 YTD)[15][97] - **Daily Rebalancing**: Excess Return: 0.75% (last week), 1.87% (March), 3.72% (2026 YTD); Absolute Return: -4.49% (last week), -7.21% (March), 6.20% (2026 YTD)[15][97]

- Model Name: Timing Radar Six-Factor Model; Model Construction Idea: The model is based on a multi-dimensional perspective to time the market, considering factors such as liquidity, economic conditions, valuation, capital flow, technical indicators, and crowding; Model Construction Process: The model selects 21 indicators from the aforementioned dimensions and categorizes them into four main categories: "Valuation Cost-Effectiveness," "Macroeconomic Fundamentals," "Capital & Trend," and "Crowding & Reversal." These indicators are then used to generate a comprehensive timing score ranging between [-1,1]; Model Evaluation: The model provides a comprehensive view of market conditions from multiple dimensions, offering a nuanced perspective on market timing[1][6][8] - Factor Name: Liquidity Direction Factor; Factor Construction Idea: This factor aims to determine the direction of current monetary policy; Factor Construction Process: The factor is calculated using the average change in central bank policy tool rates and short-term market rates over the past 90 days. If the factor is greater than 0, it indicates a loose monetary policy, and if less than 0, it indicates a tight monetary policy; Current View: This week, the liquidity direction factor is greater than 0, signaling a bullish signal with a score of 1[10][12] - Factor Name: Liquidity Strength Factor; Factor Construction Idea: This factor is based on the "interest rate corridor" concept to measure the deviation of short-term market rates from policy rates; Factor Construction Process: The factor is calculated as the deviation = DR007/7-year reverse repo rate - 1, smoothed and z-scored to form the liquidity strength factor. If the factor is less than -1.5 standard deviations, it indicates a loose environment for the next 120 trading days with a score of 1, and if greater than 1.5 standard deviations, it indicates a tight environment with a score of -1; Current View: This week, the liquidity strength factor score is 0, indicating a neutral signal[13][15] - Factor Name: Credit Direction Factor; Factor Construction Idea: This factor measures the tightness of credit transmission from commercial banks to the real economy; Factor Construction Process: The factor is calculated using the monthly value of medium and long-term loans, the incremental value over the past 12 months, and the year-over-year change. If the factor rises compared to three months ago, it signals a bullish view with a score of 1, and if it falls, it signals a bearish view with a score of -1; Current View: This week, the credit direction factor is identified as an upward trend, signaling a bullish view with a score of 1[16][18] - Factor Name: Credit Strength Factor; Factor Construction Idea: This factor captures whether credit indicators significantly exceed or fall short of expectations; Factor Construction Process: The factor is calculated as (new RMB loans for the month - median expectation) / standard deviation of expectations. If the factor is greater than 1.5 standard deviations, it indicates a significantly above-expectation environment for the next 60 trading days with a score of 1, and if less than -1.5 standard deviations, it indicates a significantly below-expectation environment with a score of -1; Current View: This week, the credit strength factor score is 0[20][22] - Factor Name: Growth Direction Factor; Factor Construction Idea: This factor is based on PMI data to measure the direction of economic growth; Factor Construction Process: The factor is calculated using the 12-month average of PMI data (including official manufacturing PMI, non-manufacturing PMI, and Caixin manufacturing PMI) and the year-over-year change. If the factor rises compared to three months ago, it signals a bullish view with a score of 1, and if it falls, it signals a bearish view with a score of -1; Current View: This week, the growth direction factor is identified as a downward trend, signaling a bearish view with a score of -1[23][25] - Factor Name: Growth Strength Factor; Factor Construction Idea: This factor captures whether economic growth indicators significantly exceed or fall short of expectations; Factor Construction Process: The factor is calculated as the PMI expectation deviation = (PMI - median expectation) / standard deviation of expectations. If the factor is greater than 1.5 standard deviations, it indicates a significantly above-expectation environment for the next 60 trading days with a score of 1, and if less than -1.5 standard deviations, it indicates a significantly below-expectation environment with a score of -1; Current View: This week, the growth strength factor score is -1, signaling a bearish view[26][28] - Factor Name: Inflation Direction Factor; Factor Construction Idea: This factor measures the current inflation level's impact on monetary policy; Factor Construction Process: The factor is calculated as 0.5 × smoothed CPI year-over-year value + 0.5 × raw PPI year-over-year value. If the factor decreases compared to three months ago, it indicates a downward inflation environment, signaling a bullish view with a score of 1, and if it increases, it signals a bearish view with a score of -1; Current View: This week, the inflation direction factor is identified as an upward trend, signaling a bearish view with a score of -1[29][31] - Factor Name: Inflation Strength Factor; Factor Construction Idea: This factor captures whether inflation indicators significantly exceed or fall short of expectations; Factor Construction Process: The factor is calculated as the average of CPI and PPI expectation deviations = (reported value - median expectation) / standard deviation of expectations. If the factor is less than -1.5, it indicates a significantly below-expectation environment for the next 60 trading days with a score of 1, and if greater than 1.5 standard deviations, it indicates a significantly above-expectation environment with a score of -1; Current View: This week, the inflation strength factor score is -1, signaling a bearish view[32][33] - Factor Name: Shiller ERP; Factor Construction Idea: This factor adjusts for economic cycle fluctuations in corporate earnings to assess current market valuation levels; Factor Construction Process: The factor is calculated using the average inflation-adjusted earnings over the past 6 years to compute the Shiller PE, then calculating Shiller ERP = 1 / Shiller PE - 10-year government bond yield, and z-scoring the past 6 years to obtain the score; Current View: This week, the Shiller ERP score is 0.05, up by 0.09 points[34][35][38] - Factor Name: PB; Factor Construction Idea: This factor follows a similar process to ERP for PB; Factor Construction Process: The factor is calculated as PB × (-1) and z-scored over the past 6 years, truncated at 1.5 standard deviations and standardized to a range of ±1; Current View: This week, the PB score is -0.57, up by 0.17 points[36][37] - Factor Name: AIAE; Factor Construction Idea: This factor reflects the overall market risk preference based on the aggregate investor allocation to equities; Factor Construction Process: The factor is calculated as AIAE = total market cap of CSI All Share Index / (total market cap of CSI All Share Index + total debt of the real economy), then AIAE × (-1) and z-scored over the past 6 years to obtain the score; Current View: This week, the AIAE score is -0.80, up by 0.20 points[39][40] - Factor Name: Margin Trading Increment; Factor Construction Idea: This factor measures market sentiment based on the source of leverage funds; Factor Construction Process: The factor is calculated as the difference between margin financing balance and margin securities balance, and the average increment over the past 120 days compared to the past 240 days. If the 120-day average increment is greater than the 240-day average increment, it signals a bullish view with a score of 1, and if less, it signals a bearish view with a score of -1; Current View: This week, the short-term margin trading increment is less than the long-term increment, signaling a bearish view with a score of -1[42][44] - Factor Name: Trading Volume Trend; Factor Construction Idea: This factor measures market trading activity and capital flow; Factor Construction Process: The factor is calculated as the log trading volume moving average distance = ma120 / ma240 - 1. If the max(10) = max(30) = max(60), it signals a bullish view with a score of 1, and if the min(10) = min(30) = min(60), it signals a bearish view with a score of -1; Current View: This week, the trading volume signal is neutral with a score of 0[45][47] - Factor Name: China Sovereign CDS Spread; Factor Construction Idea: This factor represents the pricing level of China's economic and sovereign credit risk by overseas investors; Factor Construction Process: The factor is calculated as the 20-day difference of the smoothed CDS spread. If the 20-day difference is less than 0, it signals a bullish view with a score of 1, and if greater than 0, it signals a bearish view with a score of -1; Current View: This week, the 20-day difference of the CDS spread is greater than 0, signaling a bearish view with a score of -1[48][50] - Factor Name: Overseas Risk Aversion Index; Factor Construction Idea: This factor captures overseas market risk preference; Factor Construction Process: The factor is calculated as the 20-day difference of the smoothed Citi RAI Index. If the 20-day difference is less than 0, it signals a bullish view with a score

- The report highlights that large-cap stocks outperformed small-cap stocks last week, and technical factors performed relatively well[1][5] - The report summarizes the performance of various quantitative stock portfolios constructed by Guotai Haitong Securities' financial engineering team for the past week, March, and 2026 year-to-date (YTD)[8] - The report provides detailed performance metrics for multiple factor portfolios, including aggressive, balanced, and enhanced index portfolios, as well as specific combinations like PB-earnings, GARP, and small-cap value and growth portfolios[9][10][11][13][15][26][28][30][33][35][37][39][40] - The report evaluates the performance of single factors, including style factors (market cap, PB, PE_TTM), technical factors (reversal, turnover rate, volatility), and fundamental factors (ROE, SUE, expected net profit adjustment)[44][45][46][50][51][53][54] - The report provides specific performance values for each factor and portfolio, such as weekly, monthly, and YTD returns, excess returns, tracking errors, and maximum relative drawdowns[9][10][11][13][15][26][28][30][33][35][37][39][40][45][46][50][51][53][54]

Quantitative Models and Construction Methods - **Model Name**: Dividend-adjusted futures basis model **Model Construction Idea**: The model adjusts the futures basis by incorporating the impact of expected dividends during the contract's lifespan[8][17] **Model Construction Process**: 1. The futures basis is defined as the difference between the futures contract closing price and the underlying index closing price[17]. 2. The adjustment formula is: $ \text{Adjusted Basis} = \text{Actual Basis} + \text{Expected Dividends during Contract's Lifespan} $[18] 3. Annualized basis calculation: $ \text{Annualized Basis} = (\text{Actual Basis} + \text{Expected Dividend Points}) / \text{Index Price} \times 360 / \text{Days Remaining in Contract} $[19] **Model Evaluation**: The model effectively accounts for dividend impacts, providing a more accurate representation of futures pricing dynamics[8][17] - **Model Name**: Continuous hedging strategy **Model Construction Idea**: This strategy leverages the convergence of futures basis over time by continuously rolling over contracts[40] **Model Construction Process**: 1. Backtesting period: July 22, 2022, to March 20, 2026[41] 2. Portfolio allocation: 70% of funds in the spot index and 30% in short futures contracts[41] 3. Rebalancing rule: Roll over contracts when fewer than two days remain until expiration, using the closing price to open new positions in the next contract[41] **Model Evaluation**: The strategy provides stable returns but is sensitive to transaction costs and market liquidity[40][41] - **Model Name**: Minimum basis strategy **Model Construction Idea**: This strategy selects futures contracts with the smallest annualized basis for hedging[42] **Model Construction Process**: 1. Backtesting period: July 22, 2022, to March 20, 2026[42] 2. Portfolio allocation: 70% of funds in the spot index and 30% in short futures contracts[42] 3. Selection rule: Calculate the annualized basis for all available contracts and choose the one with the smallest basis[42] 4. Holding period: Hold the selected contract for eight trading days or until fewer than two days remain until expiration[42] **Model Evaluation**: The strategy reduces basis risk but requires frequent rebalancing, increasing operational complexity[42] Model Backtesting Results - **Dividend-adjusted futures basis model**: - IC contract: Current basis -7.30%, weekly high -5.81%[20] - IF contract: Current basis -4.51%, weekly high -4.23%[25] - IH contract: Current basis -0.54%, weekly high 0.26%[30] - IM contract: Current basis -8.91%, weekly low -9.69%[35] - **Continuous hedging strategy**: - IC: Annualized return -3.27%, volatility 3.82%, max drawdown -12.10%, net value 0.8862[44] - IF: Annualized return 0.25%, volatility 2.85%, max drawdown -3.95%, net value 1.0093[49] - IH: Annualized return 0.99%, volatility 2.89%, max drawdown -4.22%, net value 1.0364[53] - IM: Annualized return -6.17%, volatility 4.30%, max drawdown -21.04%, net value 0.7936[55] - **Minimum basis strategy**: - IC: Annualized return -1.66%, volatility 4.55%, max drawdown -8.56%, net value 0.9412[44] - IF: Annualized return 1.16%, volatility 2.99%, max drawdown -4.06%, net value 1.0428[49] - IH: Annualized return 1.62%, volatility 2.94%, max drawdown -3.91%, net value 1.0599[53] - IM: Annualized return -3.95%, volatility 5.13%, max drawdown -14.41%, net value 0.8638[55] Quantitative Factors and Construction Methods - **Factor Name**: VIX (Volatility Index) **Factor Construction Idea**: Reflects market expectations of future volatility based on option prices[58] **Factor Construction Process**: 1. Adjusted methodology based on international practices and domestic market characteristics[58] 2. Incorporates term structure to show volatility expectations across different time horizons[58] **Factor Evaluation**: Provides valuable insights into market sentiment and risk levels[57][58] - **Factor Name**: SKEW (Skewness Index) **Factor Construction Idea**: Measures the asymmetry in implied volatility across different strike prices, indicating market sentiment towards extreme events[62] **Factor Construction Process**: 1. Tracks the slope of implied volatility curves for options with varying strike prices[62] 2. Higher values indicate increased concern about downside risks, while lower values suggest reduced tail risk[62] **Factor Evaluation**: Useful for assessing market expectations of tail risks and extreme events[62] Factor Backtesting Results - **VIX**: - 30-day VIX levels: - SSE 50: 21.61 (74% percentile since 2024)[58] - CSI 300: 20.97 (69% percentile since 2024)[58] - CSI 500: 34.19 (91% percentile since 2024)[58] - CSI 1000: 28.52 (65% percentile since 2024)[58] - **SKEW**: - 30-day SKEW levels: - SSE 50: 100.79 (72.2% percentile since 2024)[63] - CSI 300: 103.11 (73.1% percentile since 2024)[63] - CSI 500: 99.94 (44.0% percentile since 2024)[63] - CSI 1000: 98.68 (13.1% percentile since 2024)[63]

Quantitative Models and Construction Methods 1. Model Name: Excellent Fund Performance Enhancement Portfolio - **Model Construction Idea**: Transition from benchmarking broad-based indices to benchmarking active equity funds, leveraging quantitative methods to enhance fund selection and achieve "best of the best"[4][19][49] - **Model Construction Process**: - Benchmark against the median return of active equity funds, represented by the biased equity hybrid fund index (885001.WI)[19][49] - Use performance stratification to select superior funds, neutralizing return-related factors to avoid style concentration[49] - Optimize the portfolio to control deviations in individual stocks, industries, and styles relative to the selected fund holdings[50] - Incorporate transaction costs and fund positions (90% in this period) into return calculations[19][49] - **Model Evaluation**: Demonstrates strong stability and the ability to consistently outperform the median of active equity funds[50] 2. Model Name: Outperformance Stock Selection Portfolio - **Model Construction Idea**: Focus on stocks with significant outperformance events, leveraging both fundamental and technical dimensions for selection[5][55] - **Model Construction Process**: - Screen stocks based on research report titles indicating outperformance and analysts' upward revisions of net profit[5][55] - Select stocks with both fundamental support and technical resonance from the outperformance stock pool[5][55] - Construct the portfolio by combining these selected stocks[55] - **Model Evaluation**: Consistently ranks in the top 30% of active equity funds annually, showcasing strong performance[56] 3. Model Name: Brokerage Golden Stock Performance Enhancement Portfolio - **Model Construction Idea**: Use the brokerage golden stock pool as the stock selection space and constraint benchmark, optimizing the portfolio to control deviations in individual stocks and styles[6][33][60] - **Model Construction Process**: - Benchmark against the biased equity hybrid fund index[33][60] - Optimize the portfolio to further refine the brokerage golden stock pool, aiming for stable outperformance of the benchmark[60] - Incorporate transaction costs and fund positions (90% in this period) into return calculations[33][60] - **Model Evaluation**: Demonstrates strong performance, consistently ranking in the top 30% of active equity funds annually[61] 4. Model Name: Growth and Stability Portfolio - **Model Construction Idea**: Focus on the timing of excess returns for growth stocks, using a "time-series first, cross-section later" approach to construct a two-dimensional evaluation system[7][38][65] - **Model Construction Process**: - Introduce an "excess return release map" to identify the strongest phases of excess return before and after positive events, such as earnings pre-announcements[65] - Prioritize stocks closer to the formal financial report disclosure date, and use multi-factor scoring to select high-quality stocks when the sample size is large[7][65] - Incorporate mechanisms like weak balance, transition, buffering, and risk avoidance to reduce turnover and manage risks[65] - **Model Evaluation**: Consistently ranks in the top 30% of active equity funds annually, with strong performance in capturing excess returns[66] --- Model Backtesting Results 1. Excellent Fund Performance Enhancement Portfolio - Annualized return (2012-2025): 21.40%[51] - Annualized excess return over biased equity hybrid fund index: 9.85%[51] - Consistently ranks in the top 30% of active equity funds annually[51] 2. Outperformance Stock Selection Portfolio - Annualized return (2010-2025): 31.11%[56] - Annualized excess return over biased equity hybrid fund index: 23.98%[56] - Consistently ranks in the top 30% of active equity funds annually[56] 3. Brokerage Golden Stock Performance Enhancement Portfolio - Annualized return (2018-2025): 21.71%[61] - Annualized excess return over biased equity hybrid fund index: 14.18%[61] - Consistently ranks in the top 30% of active equity funds annually[61] 4. Growth and Stability Portfolio - Annualized return (2012-2025): 36.34%[66] - Annualized excess return over biased equity hybrid fund index: 26.33%[66] - Consistently ranks in the top 30% of active equity funds annually[66]

Quantitative Models and Construction Methods - **Model Name**: CRR Binomial Tree Pricing Model **Model Construction Idea**: The model incorporates embedded clauses, credit spreads, and other factors to improve pricing accuracy compared to traditional methods like BSM[13][40] **Model Construction Process**: 1. Define the theoretical value of convertible bonds using the CRR binomial tree model. 2. Calculate the "pricing deviation" as the difference between the CRR theoretical pricing and the market price. 3. Use this deviation to assess whether the convertible bond is undervalued or overvalued. **Evaluation**: The model is more precise in pricing convertible bonds due to its consideration of embedded clauses and credit spreads[13][40] - **Model Name**: Convertible Bond Fund Delta Tracking Model **Model Construction Idea**: Dynamically track the style allocation of convertible bond funds and calculate weighted DELTA values to observe fund behavior[38][40] **Model Construction Process**: 1. Define convertible bond funds as funds primarily investing in convertible bonds. 2. Use regression models to dynamically track the style allocation proportions of convertible bond holdings. 3. Calculate weighted DELTA values based on style indices to monitor fund linkage with equity market movements. **Evaluation**: Provides insights into fund behavior and style shifts, with DELTA values reflecting the linkage between fund performance and equity market trends[38][40] Model Backtesting Results - **CRR Binomial Tree Pricing Model**: - Pricing deviation median: -12.79 yuan - Weighted deviation: -9.67 yuan[14][16][40] - **Convertible Bond Fund Delta Tracking Model**: - Average DELTA value: 72.41% - DELTA contribution ratio: 7:2:1 (equity, balanced, debt styles)[38][40] Quantitative Factors and Construction Methods - **Factor Name**: Convertible Bond Low Valuation Momentum Factor **Factor Construction Idea**: Identify convertible bonds with market prices lower than CRR theoretical pricing and positive short-term momentum in underlying stocks[40][42] **Factor Construction Process**: 1. Initial screening: - Bond rating AA- or above - Outstanding balance ≥ 2 billion yuan - Non-ST stocks historically - No rating downgrades or negative outlooks historically - Major shareholder pledge ratio < 90% - Recent 10-day trading activity - Redemption progress < 5 days - Not below debt floor 2. Bond classification: - Divide bonds into equity, balanced, and debt styles based on parity levels (<90, 90-110, >110). 3. Selection criteria: - Comprehensive scoring based on valuation (absolute and relative pricing) and short-term stock momentum. - Select top 10 bonds from each style for a total of 30 bonds. 4. Weighting and rebalancing: Equal weighting, monthly rebalancing[40][42] **Evaluation**: The factor effectively identifies undervalued convertible bonds with potential for positive returns[40][42] Factor Backtesting Results - **Convertible Bond Low Valuation Momentum Factor**: - Monthly return: 1.21% - Long-term annualized return (since 2017): 16.59% - Maximum drawdown: 11.26% - Return-to-drawdown ratio: 1.47 - Monthly win rate: 67.35%[40][46]

- The report introduces the dividend point estimation method for stock indices, emphasizing its importance in accurately calculating the premium or discount of stock index futures contracts, which track price indices rather than total return indices[42][11][43] - The dividend point estimation formula is presented as follows: $$\text{Dividend Points} = \sum_{n=1}^{N} \frac{\text{Dividend Amount of Component Stock}}{\text{Total Market Value of Component Stock}} \times \text{Component Stock Weight} \times \text{Index Closing Price}$$ This formula requires the dividend date of component stocks to be greater than the current date and less than or equal to the futures contract expiration date[42][43][45] - The estimation process involves several steps: 1. Determining whether the company has disclosed dividend amounts and dates. If not, estimating these values based on historical data[45] 2. Estimating dividend amounts using the formula: $$\text{Dividend Amount} = \text{Net Profit} \times \text{Dividend Payout Ratio}$$ Net profit is predicted dynamically based on historical profit distributions, while the dividend payout ratio is estimated using historical averages[50][52][53] 3. Predicting the ex-dividend date using a linear extrapolation method based on historical intervals between dividend announcements and ex-dividend dates[54][59][57] - The report highlights the accuracy of the dividend point estimation model, showing that for indices like the SSE 50 and CSI 300, the annual prediction error is within 5 points, while for CSI 500 and CSI 1000, the error is around 10 points[60][64][66] - The report also tracks the premium or discount levels of major stock index futures contracts (IH, IF, IC, IM), analyzing their historical trends and current positions in historical percentiles. For example, IH is at the 23rd percentile, IF at the 34th, IC at the 64th, and IM at the 57th percentile[27][29][34]

Quantitative Models and Construction Methods 1. Model Name: Industry Consensus Expectation Dispersion Indicator - **Model Construction Idea**: The model is based on the hypothesis that industries with higher dispersion in analyst consensus expectations are more likely to deliver earnings surprises, which can lead to excess returns when market expectations shift from divergence to consensus[2][11]. - **Model Construction Process**: 1. For each stock, calculate the dispersion of FY1 net profit forecasts using the formula: $ \text{Dispersion} = \frac{\text{Max(FY1 Net Profit Forecast)} - \text{Min(FY1 Net Profit Forecast)}}{|\text{Median(FY1 Net Profit Forecast)}|} $ This measures the degree of disagreement among analysts regarding a stock's net profit forecast[12]. 2. To mitigate the impact of outliers, apply a winsorization process to the dispersion values within each industry[12]. 3. Aggregate the stock-level dispersion values to the industry level using weighted averages based on stock weights[12]. 4. Adjust for inherent differences in profit volatility across industries by calculating the rolling 2-year percentile rank of the dispersion indicator for each industry. This ensures cross-industry comparability[14]. - **Model Evaluation**: The model effectively identifies industries with higher potential for earnings surprises, particularly during the April earnings season, as evidenced by its historical performance[15]. --- Model Backtesting Results 1. Industry Consensus Expectation Dispersion Indicator - **April Performance**: - Median excess return for the long portfolio (industries with the highest dispersion): 2%[15]. - Excess return win rate: 88% (2018-2025 period)[15]. - **Overall Performance**: - Industries with high dispersion consistently outperformed the equal-weighted benchmark, while industries with low dispersion underperformed[15]. - **Latest Results (as of February 2026)**: - Industries with the highest dispersion: Retail, Communication, Food & Beverage[18]. - Industries with the lowest dispersion: Basic Chemicals, Non-Ferrous Metals, Building Materials[18].