怎样从历史走势规律发现ETF投资机会

Quantitative Models and Construction Methods 1. Model Name: Position Parameter Table - Model Construction Idea: The model aims to predict future index returns by identifying the relationship between historical price-volume patterns and future performance. It establishes a multi-to-one mapping between historical periods and future periods to construct a position parameter table[40][49]. - Model Construction Process: 1. Fix the future return period to 20 trading days for monthly portfolio construction[40]. 2. Traverse historical price-volume periods ranging from 5 to 240 days to identify the most effective historical period for prediction[40]. 3. Use polar coordinates to represent index states, where the radial distance (ρ) is calculated using the Mahalanobis distance: $ \rho = \sqrt{(x-y)^T \cdot \Sigma^{-1} \cdot (x-y)} $ Here, $x$ represents the current price and volume, $y$ represents historical price and volume, and $Σ$ is the covariance matrix[14][15]. 4. The polar angle (θ) is calculated using the arctangent function: $ \theta = \arctan2(\text{Volume Change}, \text{Price Change}) $[15]. 5. Divide the polar coordinate plane into 80 regions by segmenting the radial distance into 5 equal parts and the polar angle into 16 equal parts[31]. 6. Map each region to the average future 20-day return of indices within that region[46]. 7. Expand the table along two dimensions: - Dimension 1: Calculate the relationship between multiple historical periods and future returns for a single historical date[49]. - Dimension 2: Calculate the relationship between multiple historical dates and future returns for a single historical period[47][49]. - Model Evaluation: The model effectively captures historical price-volume patterns and their relationship with future returns, enabling the construction of robust ETF portfolios[40][49]. 2. Model Name: ETF Selection Based on Position Parameter Table - Model Construction Idea: This model selects ETFs by matching historical price-volume patterns with regions in the position parameter table that correspond to high future returns[51]. - Model Construction Process: 1. At each rebalancing point, calculate the price-volume patterns of indices over multiple historical periods[51]. 2. Match these patterns with the position parameter table to identify regions with the highest future return rankings[51]. 3. Select indices within these optimal regions to construct the ETF portfolio[51]. - Model Evaluation: The model demonstrates strong predictive power and adaptability, as evidenced by its performance across different time windows[51][53]. --- Model Backtesting Results 1. Position Parameter Table - Annualized Return: 17.27% (Full Sample), 14.65% (Fixed Window), 17.45% (Rolling Window)[65] - Cumulative Return: 351.83% (Full Sample), 264.93% (Fixed Window), 358.64% (Rolling Window)[65] - Excess Return (2021-2025/6/20): - Fixed Window: 35.81% (ETF Portfolio: +15.82%, Equal-Weighted ETF: -14.72%)[57] - Rolling Window: 77.51% (ETF Portfolio: +51.38%, Equal-Weighted ETF: -14.72%)[60] 2. ETF Selection Based on Position Parameter Table - Annual Returns: - 2021: 6.11% (Fixed Window), 15.96% (Rolling Window), 13.31% (Equal-Weighted ETF)[65] - 2022: -9.28% (Fixed Window), -5.13% (Rolling Window), -22.15% (Equal-Weighted ETF)[65] - 2023: -4.51% (Fixed Window), -6.88% (Rolling Window), -8.35% (Equal-Weighted ETF)[65] - 2024: 16.03% (Fixed Window), 29.12% (Rolling Window), 7.18% (Equal-Weighted ETF)[65] --- Quantitative Factors and Construction Methods 1. Factor Name: Price-Volume Pattern (Polar Coordinates) - Factor Construction Idea: This factor quantifies the price-volume relationship of indices using polar coordinates to identify distinct movement patterns[15][19]. - Factor Construction Process: 1. Represent price and volume changes in a Cartesian coordinate system, dividing them into four quadrants: - Quadrant 1: Price up, Volume up (0°-90°) - Quadrant 2: Price down, Volume up (90°-180°) - Quadrant 3: Price down, Volume down (180°-270°) - Quadrant 4: Price up, Volume down (270°-360°)[19][20]. 2. Transition to polar coordinates, where the radial distance (ρ) measures the magnitude of change, and the polar angle (θ) indicates the direction of change[15][19]. 3. Use historical data to map price-volume patterns to future returns, identifying regions with high predictive power[28][31]. - Factor Evaluation: The factor provides a precise and intuitive representation of price-volume dynamics, enabling effective prediction of future index performance[19][31]. --- Factor Backtesting Results 1. Price-Volume Pattern (Polar Coordinates) - Future 20-Day Return by Region: - Regions with high radial distances and specific angular ranges (e.g., 0°-90° for strong upward momentum) exhibit higher average returns[28][31]. - Historical data shows that indices in regions corresponding to "Volume Up + Price Up" (0°-90°) and "Volume Down + Price Down" (180°-270°) tend to perform better in the future[25][28].