Quantitative Models and Construction Methods - **Model Name**: DTW Timing Model **Model Construction Idea**: Based on similarity analysis, the model evaluates the similarity between current index trends and historical trends to predict future market movements[22][24] **Model Construction Process**: 1. Identify historical market segments with high similarity to the current market using DTW (Dynamic Time Warping) distance instead of Euclidean distance to avoid sequence mismatches[24][26] 2. Calculate the weighted average future returns and standard deviation of the selected historical segments (weights are the inverse of the DTW distance)[22] 3. Generate trading signals based on the average future returns and standard deviation thresholds[22] **Formula**: $ Weighted\ Average\ Return = \frac{\sum_{i=1}^{n} \frac{Return_i}{DTW\ Distance_i}}{\sum_{i=1}^{n} \frac{1}{DTW\ Distance_i}} $ $ Weighted\ Standard\ Deviation = \sqrt{\frac{\sum_{i=1}^{n} \frac{(Return_i - Weighted\ Average\ Return)^2}{DTW\ Distance_i}}{\sum_{i=1}^{n} \frac{1}{DTW\ Distance_i}}} $ **Evaluation**: The DTW algorithm improves timing accuracy by addressing sequence mismatches, and the model performs well under normal market conditions but struggles with sudden macroeconomic changes[9][24][26] - **Model Name**: Foreign Capital Timing Model **Model Construction Idea**: Extract timing signals from price movements of foreign-listed assets related to A-shares[32][33] **Model Construction Process**: 1. Use FTSE China A50 futures (Singapore market) to construct premium/discount and price divergence indicators[32] 2. Use Southern A50 ETF (Hong Kong market) to construct price divergence indicators[32] 3. Combine signals from both assets to form the final foreign capital timing signal[32] **Evaluation**: The model effectively captures foreign capital dynamics and provides stable timing signals, with high annualized returns and favorable risk metrics[15][32] --- Model Backtesting Results - **DTW Timing Model**: - Absolute return since November 2022: 34.75%[9] - Maximum drawdown: 21.36%[4] - 2024 sample performance: 15 signal switches, stable excess returns[12][14] - **Foreign Capital Timing Model**: - Annualized return (2014-2024): 18.96% (long-short), 14.19% (long-only)[15] - Maximum drawdown: 25.69% (long-short), 17.27% (long-only)[15] - Win rate: ~55%, profit-loss ratio > 2.5[15] - 2024 sample performance: Absolute return 31.33%, maximum drawdown 8.23%[18] --- Quantitative Factors and Construction Methods - **Factor Name**: DTW Similarity Factor **Factor Construction Idea**: Measure the similarity between current and historical market trends using DTW distance[22][24] **Factor Construction Process**: 1. Apply DTW distance to compare time-series data of market indices[24] 2. Select historical segments with high similarity for prediction[22] 3. Calculate weighted future returns and standard deviation based on DTW distance[22] **Formula**: $ DTW\ Distance = \min\{ \sum_{i=1}^{n} |x_i - y_j| + \text{Boundary\ Constraints} \} $ **Evaluation**: DTW-based factors outperform traditional Euclidean distance methods in time-series analysis, providing more accurate predictions[24][26] - **Factor Name**: Foreign Capital Premium/Divergence Factor **Factor Construction Idea**: Extract timing signals from foreign-listed assets related to A-shares[32][33] **Factor Construction Process**: 1. Calculate premium/discount and divergence metrics for FTSE China A50 futures and Southern A50 ETF[32] 2. Combine metrics to form composite timing signals[32] **Evaluation**: The factor effectively captures foreign capital dynamics and provides robust timing signals[32][33] --- Factor Backtesting Results - **DTW Similarity Factor**: - Predictive accuracy improves with DTW distance, outperforming traditional methods in time-series analysis[24][26] - **Foreign Capital Premium/Divergence Factor**: - High annualized returns and favorable risk metrics, with consistent performance across different market conditions[15][32]