Group 1: Solution to the Declining Performance of Value Strategies - The report highlights that traditional price multiple-based valuation methods have struggled to accurately predict stock returns, leading to a decline in value strategy performance [2][8] - A new industry valuation model based on residual income is proposed, which calculates intrinsic value (IV) as the sum of equity book value and the present value of future economic profits [2][8] - The intrinsic value to market value ratio (IVM) has been shown to effectively predict stock returns, with a long/short portfolio based on IVM yielding significant CAPM alpha returns from 1999 to 2023, even after accounting for transaction costs [2][8] Group 2: Alternative Asset Allocation to Bonds - The report identifies a need for new investment options as the traditional 60/40 stock-bond allocation fails to provide adequate risk protection during stock market downturns [3][9] - The analysis of U.S. company data from 1975 to 2021 reveals low-risk "safe equities" that exhibit relatively low future earnings risk, which can provide protection during market declines while also having potential upside during market increases [3][9] - Safe equities are positively correlated with the stock market and negatively correlated with bonds, offering higher long-term returns compared to bonds, making them a suitable alternative for investors seeking to minimize losses during downturns [3][9] Group 3: Measuring Long-Term Investor Returns - The report discusses the inadequacies of existing methods for measuring long-term investment returns, particularly in relation to transaction strategies, cash flow reinvestment, and consumption [4][11] - An analysis of over 71,000 global stocks suggests that arithmetic mean returns are unsuitable for long-term investors, while geometric mean returns apply only to specific buy-and-hold strategies [4][11] - The report introduces the concept of sustainable returns, defined as net holding returns divided by total holding returns, which reflects the proportion of fixed consumption that can be supported by the investment, providing a new perspective on evaluating long-term investment performance [4][11]

Group 1: Industry Insights - The coal industry is experiencing a continued decline in prosperity, while the cement and steel industries are expected to see positive profit growth year-on-year. Conversely, the coal and glass industries are projected to have negative profit growth [1] - The automotive electronics sector is poised for a turning point, with the rise of intelligent driving and the expansion of affordable smart technology, particularly with companies like BYD leading the charge [2] - The petrochemical industry is highlighted for its strategic importance in energy and food security, with state-owned enterprises expected to play a crucial role in ensuring supply amidst geopolitical tensions [4] Group 2: Company-Specific Analysis - Yuexiu Property is projected to achieve a revenue of 86.4 billion yuan in 2024, reflecting a year-on-year growth of 7.7%, despite a significant decline in net profit due to reduced gross margins [8] - Zhongxin Fluorine Materials is facing pressure on its performance due to declining prices of its pharmaceutical and agricultural intermediate products, alongside high depreciation costs from new capacity [9] - China Petroleum's major shareholder plans to increase its stake in the company, with expected net profits of 173 billion yuan, 178.4 billion yuan, and 182.9 billion yuan for the years 2025 to 2027 [10]

Group 1 - The report introduces a quantitative model solution for addressing the value-growth style switching issue based on odds and win rates [1][8] - The overall market growth style portfolio had a return of -0.55%, while the value style portfolio had a return of 0.18% in the last week [8] Group 2 - The estimated odds for the growth style is 1.01, while for the value style it is 1.03, indicating a negative correlation between relative valuation levels and expected odds [2][14] - The current win rate for the growth style is 31.12%, and for the value style, it is 68.88%, based on seven win rate indicators [3][17] Group 3 - The latest investment expectations calculated show a growth style expectation of -0.38 and a value style expectation of 0.40, leading to a recommendation for the value style [4][18] - Since 2013, the annualized return of the style rotation model based on investment expectations is 26.73%, with a Sharpe ratio of 0.98 [4][19]

Quantitative Models and Construction Methods 1. Model Name: Interest Rate Price-Volume Multi-Cycle Timing Strategy - **Model Construction Idea**: This model uses kernel regression algorithms to identify the trend patterns of interest rates, capturing support and resistance levels. It integrates signals from long, medium, and short investment cycles to form a composite timing strategy[11][23] - **Model Construction Process**: 1. **Signal Generation**: - Use kernel regression to identify support and resistance levels for interest rate data across different cycles (long, medium, short)[11] - Signals are generated based on whether the interest rate breaks through these levels in an upward or downward direction[11] 2. **Cycle Frequency**: - Long cycle: Monthly signal switching - Medium cycle: Bi-weekly signal switching - Short cycle: Weekly signal switching[11] 3. **Composite Signal Scoring**: - If at least two out of three cycles show a downward breakthrough, the signal is "bullish" - If at least two out of three cycles show an upward breakthrough, the signal is "bearish"[11][23] 4. **Portfolio Construction**: - Full allocation to long-duration bonds when at least two cycles show a downward breakthrough and the trend is not upward - 50% allocation to medium-duration bonds and 50% to long-duration bonds when at least two cycles show a downward breakthrough but the trend is upward - Full allocation to short-duration bonds when at least two cycles show an upward breakthrough and the trend is not downward - 50% allocation to medium-duration bonds and 50% to short-duration bonds when at least two cycles show an upward breakthrough but the trend is downward - Equal allocation across short, medium, and long durations in other cases[23] 5. **Stop-Loss Mechanism**: - Adjust holdings to equal allocation when the daily excess return of the portfolio falls below -0.5%[23] 6. **Benchmark**: - Equal-duration strategy: 1/3 allocation to short, medium, and long durations[23] 2. Model Name: Public Bond Fund Duration and Divergence Tracking - **Model Construction Idea**: This model uses an improved regression model to dynamically track the weekly changes in the duration and divergence of public bond funds[13] - **Model Construction Process**: 1. **Duration Calculation**: - Median, 4-week moving average, and mean values of the duration (including leverage) of medium- and long-term pure bond funds are calculated[13][20] 2. **Divergence Measurement**: - Cross-sectional standard deviation of fund durations is used to measure divergence[14] 3. **Yield-to-Maturity (YTM) Analysis**: - Median, 4-week moving average, and mean values of YTM (including leverage) are calculated for the funds[20] --- Model Backtesting Results 1. Interest Rate Price-Volume Multi-Cycle Timing Strategy - **Long-Term Performance (2007.12.31 to Latest Report Date)**: - Annualized Return: 6.3% - Maximum Drawdown: 1.55% - Return-to-Drawdown Ratio: 2 - Excess Return: 1.78% - Excess Return-to-Drawdown Ratio: 0.92[23][24] - **Short-Term Performance (Since 2023 Year-End)**: - Annualized Return: 8.05% - Maximum Drawdown: 1.62% - Return-to-Drawdown Ratio: 6.91 - Excess Return: 2.78% - Excess Return-to-Drawdown Ratio: 2.85[4][23][24] - **Historical Success Rates (18 Years)**: - Absolute Return > 0: 100% - Excess Return > 0: 100%[24] 2. Public Bond Fund Duration and Divergence Tracking - **Duration Metrics**: - Median Duration: 3.13 years - 4-Week Moving Average: 3.19 years - Mean Duration: 3.4 years - Historical 5-Year Percentile: 91.51%[13][14] - **Divergence Metrics**: - Cross-Sectional Standard Deviation: 2.03 years - Historical 5-Year Percentile: 98.46%[14] - **YTM Metrics**: - Median YTM: 1.99% - 4-Week Moving Average: 2.12% - Mean YTM: 2.1%[20]

Quantitative Models and Construction Methods - **Model Name**: Dividend Impact Prediction Model **Model Construction Idea**: The model aims to predict the impact of dividends on stock index futures contracts by estimating the dividend points and their influence on futures pricing. It incorporates historical dividend patterns, company financials, and theoretical pricing models to derive the expected impact [9][12][24] **Model Construction Process**: 1. **Estimate Component Stocks' Net Profit**: Use annual reports, financial forecasts, and other available data to estimate the net profit of index component stocks [27][28] 2. **Calculate Pre-Tax Dividend Total**: Based on the assumption that the dividend payout ratio remains unchanged, calculate the total pre-tax dividend for each stock as: - **Pre-Tax Dividend Total** = Estimated Net Profit × Dividend Payout Ratio [32] 3. **Calculate Dividend Impact on Index**: - **Dividend Yield** = Tax-Adjusted Dividend Total / Latest Market Value - **Dividend Points Impact** = Stock Weight × Dividend Yield - Adjust stock weights using the formula: $$\mathrm{w_{it}={\frac{w_{i0}\times\mathrm{\scriptsize{\left(\mathrm{\scriptsize{\normalsize1+R\mathrm{\scriptsize{\normalsize1}}}\right)}}}{\sum_{1}^{n}w_{i0}\times\mathrm{\scriptsize{\left(\mathrm{\scriptsize{\normalsize1+R\mathrm{\scriptsize{\normalsize1}}}\right)}}}}}}$$ where \(w_{i0}\) is the initial weight, and \(R\) is the stock's return over the period [29] 4. **Predict Impact on Futures Contracts**: Aggregate the dividend impacts for all component stocks and adjust for contract-specific factors such as ex-dividend dates and settlement timelines [30][34] **Model Evaluation**: The model is robust in incorporating historical dividend patterns and financial forecasts, but its accuracy depends on the reliability of input data and assumptions [12][27][32] - **Model Name**: Theoretical Pricing Model for Futures **Model Construction Idea**: This model calculates the theoretical price of stock index futures by accounting for dividends and risk-free rates under no-arbitrage conditions [36][37] **Model Construction Process**: 1. **Discrete Dividend Distribution**: - For \(m\) discrete dividend payments during the contract period, the present value of dividends is: $$\mathbf{D}=\sum_{\mathrm{i=1}}^{\mathrm{m}}\mathbf{D}_{\mathrm{i}}\,/(1+\phi)$$ where \(\phi\) is the risk-free rate for the period between dividend payments [36] - The theoretical futures price is then: $$F_t = (S_t - D)(1 + r)$$ where \(S_t\) is the spot price, \(D\) is the present value of dividends, and \(r\) is the risk-free rate [36] 2. **Continuous Dividend Distribution**: - For continuous dividend yields, the theoretical price is: $$F_t = S_t e^{(r-d)(T-t)}$$ where \(d\) is the annualized dividend yield, and \(r\) is the annualized risk-free rate [37] **Model Evaluation**: The model provides a theoretical benchmark for futures pricing, but deviations may occur due to market frictions and unexpected dividend changes [36][37] Model Backtesting Results - **Dividend Impact Prediction Model**: - **April Contracts**: Dividend points for major indices are as follows: - **SSE 50**: 0.00 - **CSI 300**: 0.26 - **CSI 500**: 0.30 - **CSI 1000**: 0.19 [12][16] - **Annualized Hedging Costs (Excluding Dividends)**: - **SSE 50**: -1.96% - **CSI 300**: -0.23% - **CSI 500**: 9.49% - **CSI 1000**: 13.60% [12][16] - **Theoretical Pricing Model for Futures**: - **SSE 50 Futures**: - **IH2504**: Dividend Points = 0.00, Actual Spread = 2.15, Dividend-Adjusted Spread = 2.15 - **IH2505**: Dividend Points = 0.34, Actual Spread = 0.35, Dividend-Adjusted Spread = 0.69 [12][13] - **CSI 300 Futures**: - **IF2504**: Dividend Points = 0.26, Actual Spread = 0.10, Dividend-Adjusted Spread = 0.36 - **IF2505**: Dividend Points = 8.42, Actual Spread = -5.70, Dividend-Adjusted Spread = 2.71 [13][14] - **CSI 500 Futures**: - **IC2504**: Dividend Points = 0.30, Actual Spread = -23.10, Dividend-Adjusted Spread = -22.81 - **IC2505**: Dividend Points = 4.69, Actual Spread = -66.10, Dividend-Adjusted Spread = -61.41 [14][15] - **CSI 1000 Futures**: - **IM2504**: Dividend Points = 0.19, Actual Spread = -34.86, Dividend-Adjusted Spread = -34.66 - **IM2505**: Dividend Points = 8.96, Actual Spread = -96.46, Dividend-Adjusted Spread = -87.50 [15][16]

Market Performance - The recent 5 trading days saw the Sci-Tech 50 Index decline by 1.29%, and the ChiNext Index drop by 1.12%, while the large-cap value index rose by 0.28% and the large-cap growth index increased by 0.04% [1] - The healthcare and agriculture sectors performed well, whereas the computer and defense industries lagged behind [1] Risk Premium Analysis - The static PE of the CSI All Share 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 [1] - As of January 19, 2024, the risk premium indicator was at 4.11%, marking the fifth occurrence since 2016 of exceeding 4% [1] Valuation Levels - As of March 28, 2025, the CSI All Share Index's PE TTM percentile was at 53%, with the SSE 50 and CSI 300 at 58% and 48% respectively, while the ChiNext Index was close to 14% [2] - The ChiNext Index's valuation is relatively low compared to historical averages [2] Long-term Market Trends - The Shenzhen 100 Index has experienced bear markets approximately every three years, followed by bull markets, with declines ranging from 40% to 45% [2] - The current adjustment cycle, which began in Q1 2021, appears to have sufficient time and space for a potential upward trend [2] Fund Flow and Trading Activity - In the last 5 trading days, ETF inflows totaled 16.2 billion yuan, while margin financing decreased by approximately 24.8 billion yuan [3] - The average daily trading volume across both markets was 1.2346 trillion yuan [3] Thematic Investment Focus - As of March 28, 2025, the recommended investment themes include construction materials and low-volatility dividend stocks [2][8]

Quantitative Models and Construction Methods - **Model Name**: Timing Radar Six-Factor Framework **Model Construction Idea**: The model evaluates equity market performance by integrating 21 indicators across six dimensions: liquidity, economic fundamentals, valuation, capital flows, technical trends, and crowding. These are further categorized into four major groups: "Valuation Cost-Effectiveness," "Macro Fundamentals," "Capital & Trend," and "Crowding & Reversal," generating a composite timing score within the range of [-1,1][1][5][7] **Model Construction Process**: 1. Select 21 indicators across six dimensions to represent market characteristics 2. Group indicators into four categories: - Valuation Cost-Effectiveness - Macro Fundamentals - Capital & Trend - Crowding & Reversal 3. Normalize the scores of each indicator to a range of [-1,1] 4. Aggregate the scores to compute a composite timing score within [-1,1][1][5][7] **Model Evaluation**: The model provides a comprehensive multi-dimensional perspective for market timing, offering insights into market trends and sentiment[1][5][7] Model Backtesting Results - **Timing Radar Six-Factor Framework**: - Composite Timing Score: -0.21 (Neutral to slightly bearish)[1][5][7] - Liquidity Score: -1.00 (Significant bearish signal)[1][7][9] - Economic Fundamentals Score: 0.00 (No significant signal)[1][7][9] - Valuation Score: -0.17 (Neutral signal)[1][7][9] - Capital & Trend Score: 0.50 (Significant bullish signal)[1][7][9] - Technical Trends Score: 0.00 (No significant signal)[1][7][9] - Crowding & Reversal Score: -0.69 (Significant bearish signal)[1][7][9] Quantitative Factors and Construction Methods Liquidity Factors - **Factor Name**: Monetary Direction Factor **Construction Idea**: Measures the direction of monetary policy using central bank policy rates and short-term market rates **Construction Process**: - Calculate the average change in policy and market rates over the past 90 days - If the factor > 0, monetary policy is deemed expansionary; if < 0, it is contractionary **Current View**: The factor is < 0, signaling a bearish outlook with a score of -1[11][13] - **Factor Name**: Monetary Intensity Factor **Construction Idea**: Based on the "interest rate corridor" concept, measures the deviation of short-term market rates from policy rates **Construction Process**: - Compute deviation = DR007/7-year reverse repo rate - 1 - Smooth and normalize using z-score - If the factor < -1.5 standard deviations, it indicates a bullish environment; if > 1.5, it is bearish **Current View**: The factor signals a bearish outlook with a score of -1[14][15][16] - **Factor Name**: Credit Direction Factor **Construction Idea**: Reflects the transmission of credit from banks to the real economy using long-term loan data **Construction Process**: - Calculate the 12-month incremental change in long-term loans - Compare the year-over-year change to three months prior - If the factor is rising, it is bullish; if falling, it is bearish **Current View**: The factor signals a bearish outlook with a score of -1[17][19] - **Factor Name**: Credit Intensity Factor **Construction Idea**: Captures whether credit metrics significantly exceed or fall short of expectations **Construction Process**: - Compute = (New RMB loans - median forecast) / forecast standard deviation - Normalize using z-score - If the factor > 1.5 standard deviations, it is bullish; if < -1.5, it is bearish **Current View**: The factor signals a bearish outlook with a score of -1[20][22] Economic Factors - **Factor Name**: Growth Direction Factor **Construction Idea**: Based on PMI data, measures the trend of economic growth **Construction Process**: - Calculate the 12-month moving average of PMI data - Compare the year-over-year change to three months prior - If the factor is rising, it is bullish; if falling, it is bearish **Current View**: The factor signals a bullish outlook with a score of 1[23][24] - **Factor Name**: Growth Intensity Factor **Construction Idea**: Captures whether economic growth metrics significantly exceed or fall short of expectations **Construction Process**: - Compute PMI surprise = (PMI - median forecast) / forecast standard deviation - Normalize using z-score - If the factor > 1.5 standard deviations, it is bullish; if < -1.5, it is bearish **Current View**: The factor signals a bearish outlook with a score of -1[25][27] - **Factor Name**: Inflation Direction Factor **Construction Idea**: Measures the trend of inflation using CPI and PPI data **Construction Process**: - Compute = 0.5 × smoothed CPI year-over-year + 0.5 × raw PPI year-over-year - Compare the change to three months prior - If the factor is falling, it is bullish; if rising, it is bearish **Current View**: The factor signals a bearish outlook with a score of -1[28][30] - **Factor Name**: Inflation Intensity Factor **Construction Idea**: Captures whether inflation metrics significantly exceed or fall short of expectations **Construction Process**: - Compute CPI and PPI surprises = (Reported value - median forecast) / forecast standard deviation - Average the two surprises to form the factor - If the factor < -1.5, it is bullish; if > 1.5, it is bearish **Current View**: The factor signals a bullish outlook with a score of 1[31][33] Valuation Factors - **Factor Name**: Shiller ERP **Construction Idea**: Adjusts for economic cycles to evaluate market valuation **Construction Process**: - Compute Shiller PE = average inflation-adjusted earnings over the past six years - Compute ERP = 1/Shiller PE - 10-year government bond yield - Normalize using z-score over the past three years **Current View**: The factor score decreased to 0.39[34][38] - **Factor Name**: PB **Construction Idea**: Similar to ERP, evaluates market valuation using price-to-book ratio **Construction Process**: - Compute PB × (-1) - Normalize using z-score over the past three years - Truncate to ±1 range **Current View**: The factor score decreased to -0.49[36][39] - **Factor Name**: AIAE **Construction Idea**: Reflects market-wide equity allocation and risk appetite **Construction Process**: - Compute AIAE = total market cap of CSI All Share Index / (total market cap + total debt) - Multiply by (-1) and normalize using z-score over the past three years **Current View**: The factor score decreased to -0.41[40][42] Capital Flow Factors - **Factor Name**: Margin Trading Increment **Construction Idea**: Measures market leverage and sentiment using margin trading data **Construction Process**: - Compute = financing balance - short selling balance - Compare the 120-day moving average increment to the 240-day moving average increment - If the short-term increment > long-term increment, it is bullish; otherwise, bearish **Current View**: The factor signals a bullish outlook with a score of 1[44][46] - **Factor Name**: Turnover Trend **Construction Idea**: Measures market activity and capital flow using turnover data **Construction Process**: - Compute log turnover moving average distance = ma120/ma240 - 1 - If the maximum of the 10, 30, and 60-day distances is positive, it is bullish; otherwise, bearish **Current View**: The factor signals a bullish outlook with a score of 1[47][49] - **Factor Name**: China Sovereign CDS Spread **Construction Idea**: Reflects foreign investors' sentiment towards China's credit risk **Construction Process**: - Compute the 20-day difference of smoothed CDS spreads - If the difference < 0, it is bullish; otherwise, bearish **Current View**: The factor signals a bullish outlook with a score of 1[50][51] - **Factor Name**: Overseas Risk Aversion Index **Construction Idea**: Captures global risk sentiment using Citi RAI Index **Construction Process**: - Compute the 20-day difference of smoothed RAI - If the difference < 0, it is bullish; otherwise, bearish **Current View**: The factor signals a bearish outlook with a score

Quantitative Models and Construction Methods - **Model Name**: Three-Dimensional Sentiment Model **Model Construction Idea**: The model observes market sentiment from three perspectives: sentiment expectation, sentiment temperature, and sentiment concentration, corresponding to high-frequency, medium-frequency, and low-frequency sentiment fluctuations respectively [7] **Model Construction Process**: 1. **Sentiment Expectation**: - **Indicator Significance**: Reflects short-term market sentiment through futures and options data. Futures basis rate and the inverse of options PCR (Put-Call Ratio) are used to measure sentiment [6][8] - **Formula**: $ \text{Futures Basis Rate} = \frac{\text{Futures Price} - \text{Spot Price}}{\text{Spot Price}} $ $ \text{Sentiment Expectation Composite Indicator} = \text{Mean Value + Principal Component Analysis} $ - **Evaluation**: Sentiment expectation rising indicates optimistic short-term market sentiment, while a decline suggests cautious sentiment [6][8] 2. **Sentiment Temperature**: - **Indicator Significance**: Quantifies market trading heat and fund activity, focusing on institutional/main funds as the core force. Uses "main fund buy-in rate" smoothed and calculated as a three-year rolling percentile [12] - **Formula**: $ \text{Main Fund Buy-in Rate} = \frac{\text{Large Buy-in Amount}}{\text{Total Market Turnover}} $ - **Evaluation**: Sentiment temperature rising indicates increased fund activity, while a decline suggests cooling sentiment [12] 3. **Sentiment Concentration**: - **Indicator Significance**: Measures the correlation of multi-assets in the A-share market. Uses the first principal component variance contribution rate of the CITIC three-level industry system index, smoothed with a rolling window [16] - **Evaluation**: Higher sentiment concentration indicates increased asset correlation, suggesting stronger emotional influence on the market. Extreme values above the warning line (0.83) may signal long-term market turning points [16] Model Backtesting Results - **Sentiment Expectation**: Current value as of February 20, 2025: 0.7696, up 31.02% from the previous week [9][22] - **Sentiment Temperature**: Current value as of February 20, 2025: 0.6952, up 4.43% from the previous week [13][22] - **Sentiment Concentration**: Current value as of February 20, 2025: 0.6884, down 2.17% from the previous week [18][22]

Quantitative Models and Factor Construction Multi-Horizon Factor - **Factor Name**: Multi-Horizon Factor - **Construction Idea**: This factor captures short-term reversal, medium-term momentum, and long-term reversal effects by analyzing moving average (MA) data across multiple time horizons [2][14][21] - **Construction Process**: - Calculate moving averages for different time horizons \( L = [3, 5, 10, 20, 30, 60, 90, 120, 180, 240, 270, 300] \) using the formula: \[ A_{j t,L} = \frac{P_{j,\,d-L+1}^{t} + \cdots + P_{j,d}^{t}}{L} \] where \( P_{j,d}^t \) represents the price of stock \( j \) at time \( t \) [21] - Standardize the moving average factor: \[ \tilde{A}_{j t,\,L} = \frac{A_{j t,\,L}}{P_{j}^{t}} \] [22] - Perform cross-sectional regression of stock returns on lagged standardized moving average factors: \[ r_{j,t} = \beta_{0,t} + \Sigma_{i}\beta_{i,t}\tilde{A}_{j t-1,L_{i}} + \epsilon_{j,t} \] [23] - Predict next-period regression coefficients by averaging the past 25 weeks' coefficients: \[ E\left[\beta_{i,\,t+1}\right] = \frac{1}{25}\,\sum_{m=1}^{25}\,\beta_{i,t+1-m} \] [24] - Use predicted coefficients and new factor values to estimate next-period returns: \[ E\left[r_{j,t+1}\right] = \Sigma_{i}\,E\left[\beta_{i,\,t+1}\right]\tilde{A}_{j t,\,L_{i}} \] [25] - Rank stocks by predicted returns and construct long-short portfolios [26] - **Evaluation**: The factor demonstrates strong predictive power for stock returns across different market segments, with positive IC values dominating [30][32] LLT Trend Factor - **Factor Name**: LLT Trend Factor - **Construction Idea**: To address the lagging sensitivity of MA, the LLT (Low-Lag Trendline) indicator is used as a replacement. LLT reduces delay and better captures momentum and reversal effects [14][76] - **Construction Process**: - LLT is calculated using a second-order linear filter with the recursive formula: \[ LLT = \begin{cases} P(T), & T=1,2 \\ (2-2\alpha)LLT(T-1) - (1-\alpha)^2LLT(T-2) + \left(\alpha-\frac{\alpha^2}{4}\right)P(T) \\ + \left(\frac{\alpha^2}{2}\right)P(T-1) - \left(\alpha-\frac{3}{4}\alpha^2\right)P(T-2), & \text{else} \end{cases} \] where \( \alpha = \frac{2}{1+N} \) and \( N \) is the smoothing parameter [76] - Replace MA with LLT in the multi-horizon factor construction process [76] - **Evaluation**: LLT-based factors outperform MA-based factors in terms of IC mean, positive IC ratio, and predictive power for asset returns [82][84] --- Backtesting Results Multi-Horizon Factor - **Annualized Return**: 25.40% [3][48] - **Annualized Volatility**: 14.12% [48] - **Maximum Drawdown**: 13.31% [48] - **IR**: 1.81 [48] LLT Trend Factor - **Annualized Return**: 29.58% [4][103] - **Annualized Volatility**: 10.46% [103] - **Maximum Drawdown**: 11.57% [103] - **IR**: 2.51 [103]