Testing China A-Share Factors with Different Training Periods

Summary

This account describes an informal attempt to test long-only factors for China A-shares using a visual strategy tool. The author selected factors discussed in a paper on anomalies in the Chinese A-share market, then adjusted the learning and simulation periods. The process illustrates how a published factor set can be translated into a candidate strategy and explored through historical tests.

The reported observations are tentative: an initial run appeared promising, adding the full year 2024 produced much weaker results, and a later setup trained on 2018–2022 and evaluated across 2023–2024 seemed more encouraging. No returns, benchmark, costs, portfolio rules, or detailed implementation are provided, so these impressions do not establish predictive performance. The author also notes that some factors require custom calculations, using same-month return averaged across the prior five years as an example. The piece is a learning log, not a controlled study; changing time windows can materially change apparent results and raises the need for out-of-sample and robustness checks.

Key ideas

• The author used factors reported in a study of China A-share anomalies to form a long-only test.
• Changing the learning and simulation periods led to noticeably different qualitative impressions.
• A five-year average of returns in the same calendar month is given as an example of a custom factor.
• The account supplies no detailed performance statistics, costs, benchmark, or strategy specification.
• Window sensitivity means the reported backtest impressions require more rigorous robustness checks.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.