A-Share Quant Research

Notes on China A-share market data and quant backtesting

Most English-language material on quantitative research assumes US equities. China A-shares behave differently in ways that quietly break backtests: a T+1 settlement rule, daily price limits that make a stock untradeable for a whole session, and three separate conventions for adjusting historical prices around dividends. Get any of those wrong and the backtest is measuring something other than what you think.

These notes come from running a live A-share research platform. Every number below is one we measured ourselves, including the ones that showed our own model was worse than we had claimed.

The platform itself is in Chinese and covers only mainland-listed A-shares. This section is research notes, not a product page — there is nothing to sign up for here.

Start here

How to Calculate Win Rate in a Backtest (And Why the Same Signals Give Two Opposite Answers)

The same batch of signals scored +0.56% and -0.28% depending on where we started the clock.

qfq vs hfq: which adjusted price should an A-share backtest use?

Same ticker, same dates, different closes six months later — that is forward-adjusted A-share data, and it quietly breaks reproducibility.

Backtest Pitfalls in China A-Share Research: Six Failure Modes, Ranked by How Hard They Are to Catch

Three backtest pitfalls are easy to catch. The three that killed our strategies were benchmark choice, cost conventions, and beta leakage.