Boutique Factor Test Results Updated, AI Financial Forecast, Social and More Stand Out

Recently, we selected a group of factors with a high ratio of long/short portfolio information across all types of factors over the past year, and we are sharing the results of the test with you, our readers.

All factors are neutralized by sector market capitalization, and our backtesting results show the weekly position adjustment results for the full history (2010-2025.3) and the most recent year (2024.3-2025.3).

After in-depth analysis and testing of the various factor categories, we have identified a number of factor categories that have performed particularly well both recently and historically. These factors not only exhibit high information ratios in long/short portfolios, but can also provide investors with significant alpha in real-world applications.Below are the specific outperforming factor categories and their characteristics:

1. AI Financial Forecast

Factor Description: Forecasts of public company profitability based on artificial intelligence techniques.

Test results: In both the most recent year and full history, the AI Profit Forecasting factors exhibit high IC averages and win rates, demonstrating their stability and accuracy in predicting stock returns. For example, the AI_DA_PS_30 factor has an IC average of 5.05%, a win rate of 64.15% and an annualized return of 51.15% in the most recent year, and an IC average of 5.36%, a win rate of 74.60% and an annualized return of 39.73% in the full history.

2、Socialization

Factor Description: Reflects the attention and discussion heat of users on social media platforms for specific brands or products.

Test result: The social factor GB_POST_NUM_DA30 has an average IC value of 2.87%, a win rate of 75.47% and an annualized return of 18.56% in the recent year; in the full history, the average IC value is 2.70%, the win rate is 73.35% and the annualized return is 17.45%.

3、Low Frequency

Factor Description: Focuses on long-term trends and volatility of fundamental data such as stock prices and volume.

Test results: Low-frequency factors such as APB_SKEW, TRVOV, RMVOL_20, etc. show high win rates and information ratios in both full history and the most recent year. For example, the RMVOL_20 factor has an average IC of 7.17%, a winning ratio of 81.13%, and an annualized return of 39.49% in the most recent year, and an average IC of 7.31%, a winning ratio of 77.45%, and an annualized return of 43.57% in the full history.