Quantitative Models and Construction Methods 1. Model Name: DTW Timing Model - **Model Construction Idea**: The model is based on a similarity approach, examining the similarity between current index trends and historical trends. It selects historical segments with high similarity as references and calculates the weighted average future returns and standard deviations of these segments to generate trading signals[26][28]. - **Model Construction Process**: 1. Use the DTW (Dynamic Time Warping) distance algorithm instead of Euclidean distance to measure similarity, as DTW can better handle time series misalignment issues[28]. 2. Select historical market segments with high similarity to the current market trend. 3. Calculate the weighted average future returns and standard deviations of the selected historical segments, where the weights are the inverse of the DTW distance. 4. Generate trading signals based on the average future returns and standard deviations[26]. 5. The model incorporates improved DTW algorithms, such as those with boundary constraints proposed by Sakoe-Chiba and Itakura, to address issues like "pathological matching" in traditional DTW algorithms[30]. - **Model Evaluation**: The DTW timing model is effective in general market conditions but may face challenges during periods of sudden macroeconomic policy changes, as observed in Q3 2023[17]. 2. Model Name: Foreign Capital Timing Model - **Model Construction Idea**: This model leverages information embedded in the price changes of two offshore assets related to A-shares: FTSE China A50 Index Futures (Singapore market) and the Southbound A50 ETF (Hong Kong market)[35]. - **Model Construction Process**: 1. Construct two indicators using FTSE China A50 Index Futures: basis and price divergence. 2. Combine these indicators to form the FTSE China A50 Index Futures timing signal. 3. Construct a price divergence indicator using the Southbound A50 ETF. 4. Combine the timing signals from the two assets to form the final foreign capital timing signal[35]. --- Model Backtesting Results 1. DTW Timing Model - **Absolute Return**: 38.56% since November 2022[17] - **Maximum Drawdown**: 21.32% since November 2022[17] - **2024 Performance on CSI 300**: - Absolute Return: 36.58% - Maximum Drawdown: 21.36% - Win Rate: 50.00% - Profit-Loss Ratio: 3.39[19] 2. Foreign Capital Timing Model - **Full Sample Performance (2014-2024)**: - Annualized Return: 18.96% (long-short), 14.19% (long-only) - Maximum Drawdown: 25.69% (long-short), 17.27% (long-only) - Daily Win Rate: ~55% - Profit-Loss Ratio: >2.5[21] - **2024 Performance (Out-of-Sample)**: - Absolute Return: 35.42% (long-only) - Maximum Drawdown: 8.23%[24] --- Quantitative Factors and Construction Methods No specific quantitative factors were explicitly mentioned in the report. --- Factor Backtesting Results No specific factor backtesting results were explicitly mentioned in the report.

- The DTW timing model is based on the principle of similarity and the DTW (Dynamic Time Warping) algorithm, which is a volume-price timing model[1][4][14] - The foreign capital timing model is constructed based on the divergence between foreign and domestic related assets, using four indicators reflecting foreign capital movements to generate timing signals for the A-share market[1][4][14] - The DTW timing model has shown an absolute return of 17.39% and an excess return of 17.83% relative to the CSI 300 since November 2022, with a maximum drawdown of 21.32% and a weekly win rate of over 60%[4][16] - The foreign capital timing model's long strategy has achieved an absolute return of 28.83% since 2024, with a maximum drawdown of 8.32%[4][23] - The DTW timing model uses the DTW distance algorithm instead of the Euclidean distance to measure similarity, as the DTW distance can better handle the misalignment of time series[29][30] - The DTW timing model has been improved by incorporating boundary constraints proposed by Sakoe-Chiba and Itakura to address the "pathological matching" problem of the traditional DTW algorithm[31][32][37] - The foreign capital timing strategy is based on two overseas listed assets related to A-shares: FTSE China A50 Index Futures (Singapore market) and Southern A50ETF (Hong Kong market), constructing timing signals through price divergence and premium/discount indicators[36][12] - The DTW timing model's performance in 2024 includes an absolute return of 15.68%, an excess return of 4.93%, a maximum drawdown of 21.36%, a trading win rate of 63.64%, and a profit-loss ratio of 2.64[18][19][20] - The foreign capital timing model's performance from December 30, 2014, to December 31, 2024, shows an annualized return of 18.96% (long-short) and 14.19% (long-only), with maximum drawdowns of 25.69% and 17.27%, respectively, and a daily win rate of nearly 55%[20][22][24]