Quantitative Factors and Models Summary Quantitative Factors and Construction Methods 1. **Factor Name**: Medium-term Momentum Factor **Construction Idea**: The factor is based on the cumulative returns of stocks over a medium-term window (t-12 to t-2 months), excluding the most recent month, to identify "winners" and "losers" for long-short strategies[11][12] **Construction Process**: - Formula: $ MOM_{i,t}^{M} = \prod_{k=t-12}^{t-2}(1 + r_{i,k}) - 1 $ Where $ r_{i,k} $ represents the return of stock $ i $ in month $ k $[12] **Evaluation**: The factor performs poorly in the A-share market, with weak monotonicity in decile grouping and low predictive power[13] 2. **Factor Name**: Short-term Momentum Factor **Construction Idea**: Similar to the medium-term momentum factor but with a shorter calculation window (t-6 to t-2 months), considering the faster style rotation in the A-share market[18] **Construction Process**: - Formula: $ MOM_{i,t}^{s} = \prod_{k=t-6}^{t-2}(1 + r_{i,k}) - 1 $ Where $ r_{i,k} $ represents the return of stock $ i $ in month $ k $[18] **Evaluation**: Shortening the calculation window does not improve the factor's performance, and it remains ineffective in the A-share market[19] 3. **Factor Name**: Directional Momentum (D-MOM) Factor **Construction Idea**: Predicts the "direction" of returns (positive or negative) rather than the magnitude, using a linear probability model (LPM) to enhance robustness and practicality[24][25] **Construction Process**: - Formula: $ r_{i,t}^{+} = \delta_{0} + \delta_{1}V_{i,t-1} + \delta_{2}r_{m,t-1} + \beta_{+}^{m}P_{i,t-1}^{m} + \beta_{+}^{w}P_{i,t-1}^{w} + \beta_{+}^{d}P_{i,t-1}^{d} + \beta_{-}^{m}N_{i,t-1}^{m} + \beta_{-}^{w}N_{i,t-1}^{w} + \beta_{-}^{d}N_{i,t-1}^{d} + u_{i,t} $ Where: - $ r_{i,t}^{+} $: Binary variable indicating positive returns - $ V_{i,t-1} $: Lagged idiosyncratic variance - $ r_{m,t-1} $: Lagged market return - $ P_{i,t-1}^{m}, P_{i,t-1}^{w}, P_{i,t-1}^{d} $: Positive return durations (monthly, weekly, daily) - $ N_{i,t-1}^{m}, N_{i,t-1}^{w}, N_{i,t-1}^{d} $: Negative return durations (monthly, weekly, daily)[25][26] **Evaluation**: The factor demonstrates strong stock selection ability in the A-share market, with better performance and lower risk of "momentum crashes" compared to traditional momentum factors[27][33] 4. **Factor Name**: Enhanced Directional Momentum (D-MOM) Factor **Construction Idea**: Builds on the D-MOM factor by incorporating individual stock's lagged returns as an additional predictor to capture more momentum/reversal characteristics[35] **Construction Process**: - Formula: $ r_{i,t}^{+} = \delta_{0} + \delta_{1}W_{i,t-1} + \delta_{2}r_{m,t-1} + \delta_{3}r_{i,t-1} + \beta_{4}^{m}P_{i,t-1}^{m} + \beta_{4}^{w}P_{i,t-1}^{w} + \beta_{4}^{d}P_{i,t-1}^{d} + \beta_{4}^{m}N_{i,t-1}^{m} + \beta_{2}^{w}N_{i,t-1}^{w} + \beta_{4}^{d}N_{i,t-1}^{d} + u_{i,t} $ Where $ r_{i,t-1} $ represents the lagged return of stock $ i $[35] **Evaluation**: The enhanced factor improves IC values, annualized returns, and decile monotonicity, particularly under market-cap-weighted schemes[36][42] Factor Backtesting Results 1. **Medium-term Momentum Factor**: - IC Mean: 0.019 - Annualized ICIR: 0.30 - IC>0 Probability: 59% - Annualized Long-Short Return: -1.00% - Annualized Long-Only Excess Return: 2.80% - Long-Short Sharpe Ratio: -0.10[13][18] 2. **Short-term Momentum Factor**: - IC Mean: 0.011 - Annualized ICIR: 0.20 - IC>0 Probability: 60% - Annualized Long-Short Return: -10.80% - Annualized Long-Only Excess Return: -3.00% - Long-Short Sharpe Ratio: -0.24[19][24] 3. **Directional Momentum (D-MOM) Factor**: - IC Mean: 0.059 - Annualized ICIR: 1.39 - IC>0 Probability: 86% - Annualized Long-Short Return: 21.40% (equal-weighted), 25.10% (market-cap-weighted) - Annualized Long-Only Excess Return: 11.50% (equal-weighted), 12.30% (market-cap-weighted) - Long-Short Sharpe Ratio: 1.66 (equal-weighted), 1.35 (market-cap-weighted)[33][35] 4. **Enhanced Directional Momentum (D-MOM) Factor**: - IC Mean: 0.064 - Annualized ICIR: 1.40 - IC>0 Probability: 86% - Annualized Long-Short Return: 25.40% (equal-weighted), 26.40% (market-cap-weighted) - Annualized Long-Only Excess Return: 15.10% (equal-weighted), 12.80% (market-cap-weighted) - Long-Short Sharpe Ratio: 1.84 (equal-weighted), 1.58 (market-cap-weighted)[41][42]

- The "Day and Night Convergence" new momentum factor is constructed based on the price-volume relationship during intraday and overnight trading periods, aiming to improve the stability of traditional momentum factors by incorporating transaction volume information[7][6][1] - The construction process involves splitting trading periods into day and night sessions, analyzing their respective price-volume relationships, and synthesizing a new momentum factor. The factor's IC mean is -0.045, annualized ICIR is -2.59, and its 10-group long-short strategy achieved an annualized return of 22.64%, IR of 2.85, monthly win rate of 83.33%, and maximum drawdown of 5.79% during the backtest period from 2014/01/01 to 2022/07/31[7][6][1] - The factor demonstrates superior stock selection ability compared to traditional momentum factors, which had an IR of 1.09, monthly win rate of 62.75%, and maximum drawdown of 20.35% during the same backtest period[6][7] - In the full A-share market (excluding Beijing Stock Exchange stocks), the "Day and Night Convergence" factor's 10-group long-short strategy achieved an annualized return of 17.93%, annualized volatility of 8.73%, IR of 2.05, monthly win rate of 76.98%, and maximum drawdown of 9.07% from February 2014 to August 2025[14][7][1] - For August 2025, the factor's 10-group long portfolio returned 9.49%, the short portfolio returned 9.58%, and the long-short strategy returned -0.10%[10][1]

Quantitative Models and Construction Methods 1. Model Name: "Day and Night Convergence" New Momentum Factor - **Model Construction Idea**: This model improves traditional momentum factors by incorporating the price-volume relationship during intraday and overnight trading periods. It aims to capture distinct characteristics and logic in these two periods to enhance the stability and effectiveness of momentum signals [6][7]. - **Model Construction Process**: 1. The trading session is divided into intraday and overnight periods. 2. The price-volume relationship is separately analyzed for each period. 3. Based on these analyses, the intraday and overnight factors are improved and synthesized into a new momentum factor [7]. - **Model Evaluation**: The model demonstrates significant stock selection ability, outperforming traditional momentum factors in terms of stability and performance [6][7]. --- Model Backtesting Results 1. "Day and Night Convergence" New Momentum Factor - **Annualized Return**: 18.15% [7][14] - **Annualized Volatility**: 8.79% [7][14] - **Information Ratio (IR)**: 2.07 [7][14] - **Monthly Win Rate**: 77.37% [7][14] - **Maximum Drawdown**: 9.07% [7][14] --- Quantitative Factors and Construction Methods 1. Factor Name: "Day and Night Convergence" New Momentum Factor - **Factor Construction Idea**: The factor is based on the price-volume relationship during intraday and overnight trading periods. It leverages the distinct characteristics of these periods to refine momentum signals [7]. - **Factor Construction Process**: 1. Divide the trading session into intraday and overnight periods. 2. Analyze the price-volume relationship for each period. 3. Construct separate momentum factors for intraday and overnight periods. 4. Synthesize the two factors into a new momentum factor [7]. - **Factor Evaluation**: The factor shows superior performance compared to traditional momentum factors, with higher stability and stock selection capability [6][7]. --- Factor Backtesting Results 1. "Day and Night Convergence" New Momentum Factor - **Annualized Return**: 22.64% [6] - **Information Ratio (IR)**: 2.85 [6] - **Monthly Win Rate**: 83.33% [6] - **Maximum Drawdown**: 5.79% [6]

证券研究报告·金融工程·金工定期报告 金工定期报告 20250604 "日与夜的殊途同归"新动量因子绩效月报 20250530 [Table_Tag] [Table_Summary] 报告要点 2025 年 06 月 04 日 证券分析师 高子剑 执业证书：S0600518010001 021-60199793 gaozj@dwzq.com.cn 证券分析师 庞格致 执业证书：S0600524090003 panggz@dwzq.com.cn 相关研究 《"日与夜的殊途同归"新动量因子绩 效月报 20250430》 2025-05-06 《成交量对动量因子的修正：日与夜 之殊途同归》 2022-08-17 东吴证券研究所 1 / 6 请务必阅读正文之后的免责声明部分 ◼ "日与夜的殊途同归"新动量因子多空对冲绩效（全市场）：2014 年 2 月至 2025 年 5 月，"日与夜的殊途同归"新动量因子在全体 A 股（剔 除北交所股票）中，10 分组多空对冲的年化收益率为 18.37%，年化波 动率为 8.79%，信息比率为 2.09，月度胜率为 77.94%，月度最大回撤率 为 9.07%。 ◼ 5 月份"日与 ...