To The Moon

P/E Ratio Percentile Factor Optimization and Strategy Construction ——Quantitative Strategy Research Analyst: Wu Qidi Practitioner Registration Number: A0190523020001 wuqidi@yd.com.cn Research Assistant: Wang Na Practitioner Registration Number: A0190125030006 wangna@yd.com.cn Investment Highlights: Ø Backtesting and IC Analysis of P/E Ratio Percentile Factor Across Different Window Periods This report continues the optimization approach from the previous article on P/E ratio factor portfolios, conducting optimized backtesting on the P/E ratio percentile factor. Regarding the time horizon, we constructed portfolios by selecting the 20 stocks with the lowest historical percentiles within the 6-month, 12-month, and 24-month periods preceding the rebalancing date for backtesting. The performance differences among P/E ratio percentile factor portfolios across different window periods were minimal. The portfolio of 20 stocks with the lowest P/E ratio percentiles based on a 12-month window demonstrated the optimal return. During the period from May 7, 2014, to December 10, 2025, it achieved an annualized return of 16.84%, an annualized volatility of 25.34%, and a maximum drawdown of -52.77%. Statistical tests on the factor's predictive ability revealed that the P/E ratio percentile across all time intervals exhibited highly stable and significant negative stock selection capability. The 12-month percentile factor showed the best overall performance, with a RankIC mean of -4.83%, a RankIC_IR of -0.84, and a probability of RankIC>0 at 20.83%. Further narrowing the portfolio scope to 10 stocks, the performance of the low P/E ratio portfolio within a 12-month window improved. From May 7, 2014, to December 10, 2025, the portfolio's annualized return increased to 20.80%, with an annualized volatility of 26.82% and a maximum drawdown of -44.58%. Comparison of the 20-stock portfolio with the lowest 12-month P/E ratio percentiles against the CSI 300 Index trend. Data Source: Wind, Yuanda Information Securities Research Institute. Ø Strategy Construction by Combining with Other Factors ① ROE Factor Backtesting with the addition of the ROE factor showed a decrease in the portfolio's annualized return compared to the single-factor approach. After applying a screening threshold of ROE greater than 5%, the portfolio achieved an annualized return of 16.99%, an annualized volatility of 26.52%, and a maximum drawdown of -45.03% during the period from May 7, 2014, to December 10, 2025. ② Number of Institutional Shareholders Factor Backtesting with the addition of the number of institutional shareholders factor revealed a nonlinear impact of institutional holdings on strategy performance, where moderate institutional attention can optimize the strategy. During the period from May 7, 2014, to December 10, 2025, the portfolio screened by the condition of having more than 5 institutional shareholders demonstrated the best performance, achieving an annualized return of 21.58%, an annualized volatility of 26.01%, and a maximum drawdown narrowing to -41.92%. However, when the screening criteria were tightened further to more than 10 or more than 15 institutional shareholders, the annualized returns of the portfolios decreased. Ø Risk Disclosure Historical performance does not indicate future results; the backtest model does not account for actual transaction cost rates; potential errors in other data statistics. I. P/E Ratio Percentile Factor Optimization 1. Grouped Backtesting of P/E Ratio Percentile Factor Across Different Time Intervals Historical percentile calculates the percentile point of a current metric value within a historical range, serving as a common time-series analysis method with clear economic significance. The historical P/E ratio percentile reflects how high or low a stock's current P/E valuation is within a historical range, quantitatively indicating whether the stock's P/E valuation is "cheap" historically. This report continues the optimization approach from the previous article on P/E ratio factor portfolios, conducting optimized backtesting on the P/E ratio percentile factor. Regarding the time horizon, we constructed portfolios by selecting the 20 stocks with the lowest historical percentiles within the 6-month, 12-month, and 24-month periods preceding the rebalancing date, excluding stocks with a P/E ratio percentile of 0 and ST stocks. Considering the timeliness of periodic reports and holiday factors, rebalancing was conducted annually on May 7 and November 7. If the date was not a trading day, it was postponed to the next trading day. The backtest results are as follows: Figure 1: Backtest of the 20-stock portfolio with the lowest 6-month P/E ratio percentiles (May 7, 2015 - Dec 10, 2025) Data Source: Wind, Yuanda Information Securities Research Institute. Figure 2: Backtest of the 20-stock portfolio with the lowest 12-month P/E ratio percentiles (May 7, 2015 - Dec 10, 2025) Data Source: Wind, Yuanda Information Securities Research Institute. Figure 3: Backtest of the 20-stock portfolio with the lowest 24-month P/E ratio percentiles (May 7, 2015 - Dec 10, 2025) Data Source: Wind, Yuanda Information Securities Research Institute. Table 1: Performance comparison of 20-stock portfolios with the lowest P/E ratio percentiles across different time intervals (May 7, 2015 - Dec 10, 2025) Data Source: Wind, Yuanda Information Securities Research Institute. From the backtest results, the performance differences among low P/E ratio percentile portfolios across different time intervals were small. From a return perspective, the 12-month low P/E ratio percentile portfolio performed best. As the time interval expanded, the volatility and drawdown of the P/E ratio percentile portfolios gradually decreased. The 24-month low P/E ratio percentile portfolio had the lowest volatility and drawdown, showing relatively the most stable performance. 2. Validity Test of P/E Ratio Percentile Factor Across Different Time Intervals We used IC tests to validate the effectiveness of the TTM P/E ratio percentile factor across different time intervals. Generally, a factor's IC value measures the correlation between the factor's exposure vector at period T and the stock return vector at period T+1, specifically defined as: In practical calculations, to enhance robustness, the Spearman rank correlation coefficient is typically used to calculate the Rank IC value. Factor evaluation is based on the Rank IC value series, primarily including: (1) The mean of the Rank IC series reflects factor significance, while the standard deviation measures factor stability; (2) IC_IR (the ratio of mean to standard deviation) assesses factor effectiveness; (3) The proportion of the Rank IC series greater than zero judges the stability of the factor's directional effect. Aligning with the rebalancing dates and frequency mentioned earlier, we used semi-annual data from May 7, 2015, to November 7, 2025, for analysis. The IC test results indicated that the P/E ratio percentile factor across different time intervals exhibited highly stable negative predictive ability. In over 80% of the observation periods, the negative relationship where low valuation corresponds to high returns held true. Among the different time intervals, the 12-month window factor performed best, with the optimal balance between IC mean and stability. Its RankIC value was -4.83%, RankIC_IR was -0.84, and the probability of RankIC>0 was 20.83%. Table 2: IC test of the P/E ratio percentile factor across different time intervals Data Source: Wind, Yuanda Information Securities Research Institute. 3. Further Optimization of the Portfolio Based on the Lowest 12-Month P/E Ratio Percentile Factor We further narrowed the stock selection scope, constructing a portfolio of the 10 stocks with the lowest historical percentiles within the 12-month period preceding the rebalancing date, excluding stocks with a P/E ratio percentile of 0 and ST stocks. Considering the timeliness of periodic reports and holiday factors, rebalancing was conducted annually on May 7 and November 7. If the date was not a trading day, it was postponed to the next trading day. The backtest results are as follows: Figure 4: Backtest of the 10-stock portfolio with the lowest 12-month P/E ratio percentiles (May 7, 2015 - Dec 10, 2025) Data Source: Wind, Yuanda Information Securities Research Institute. Table 3: Performance comparison of 10-stock vs. 20-stock portfolios with the lowest 12-month P/E ratio percentiles (May 7, 2015 - Dec 10, 2025) Data Source: Wind, Yuanda Information Securities Research Institute. From the backtest results, the 10-stock portfolio with the lowest 12-month P/E ratio percentiles performed better. From May 7, 2015, to December 10, 2025, the portfolio achieved an annualized return of 20.80%, an annualized volatility of 26.82%, and a maximum drawdown of -44.58%. II. Strategy Construction by Combining with Other Factors (1) ROE Factor Based on selecting the 12-month P/E ratio percentile factor, we referenced other market strategies related to low P/E ratio factor index construction, such as the CSI Profit Valuation Strategy Index, and superimposed the profitability quality factor ROE for further portfolio screening. Table 4: CSI Profit Valuation Strategy Index Data Source: Wind, Yuanda Information Securities Research Institute. We constructed a portfolio of the 10 stocks with the lowest historical percentiles within the 12-month period preceding the rebalancing date, excluding stocks with a P/E ratio percentile of 0, ST stocks, and stocks with ROE below the corresponding standard. Considering the timeliness of periodic reports and holiday factors, rebalancing was conducted annually on May 7 and November 7. If the date was not a trading day, it was postponed to the next trading day. The backtest results are as follows: Figure 5: Portfolio with ROE>5% and the lowest 12-month P/E ratio percentiles (10 stocks) shows optimal backtest performance (May 7, 2015 - Dec 10, 2025) Data Source: Wind, Yuanda Information Securities Research Institute. Table 5: Performance of the 10-stock portfolio with the lowest 12-month P/E ratio percentiles screened using the ROE factor (May 7, 2015 - Dec 10, 2025) Data Source: Wind, Yuanda Information Securities Research Institute. From the backtest results, the annualized return of the portfolio decreased after superimposing the ROE factor. After applying the screening threshold of ROE greater than 5%, the portfolio achieved an annualized return of 16.99%, an annualized volatility of 26.52%, and a maximum drawdown of -45.03% during the period from May 7, 2014, to December 10, 2025. (2) Number of Institutional Shareholders The number of institutional shareholders reflects, to some extent, the actual recognition of the company by institutional investors and provides a certain level of liquidity assurance. We constructed a portfolio of the 10 stocks with the lowest historical percentiles within the 12-month period preceding the rebalancing date, excluding stocks with a P/E ratio percentile of 0 and ST stocks, and sequentially excluding stocks with the number of institutional shareholders below the corresponding standard. Considering the timeliness of periodic reports and holiday factors, rebalancing was conducted annually on May 7 and November 7. If the date was not a trading day, it was postponed to the next trading day. The two backtest results are as follows: Figure 6: Portfolio with >5 institutional shareholders and the lowest 12-month P/E ratio percentiles (10 stocks) shows optimal backtest performance (May 7, 2015 - Dec 10, 2025) Data Source: Wind, Yuanda Information Securities Research Institute. Table 6: Performance of the 10-stock portfolio with the lowest 12-month P/E ratio percentiles screened using the number of institutional shareholders factor (May 7, 2015 - Dec 10, 2025) Data Source: Wind, Yuanda Information Securities Research Institute. From the backtest results, institutional holdings have a significant nonlinear impact on strategy performance, where moderate institutional attention can optimize the strategy. Specifically, the portfolio screened by the condition of having more than 5 institutional shareholders demonstrated the best performance, achieving an annualized return of 21.58%, an annualized volatility of 26.01%, and a maximum drawdown narrowing to -41.92%. However, when the screening criteria were tightened to more than 10 or more than 15 institutional shareholders, the annualized returns of the portfolios decreased. III. Risk Disclosure Historical performance does not indicate future results; The backtest model does not account for actual transaction cost rates; Potential errors in other data statistics.