China stock prediction: the honest numbers and why it's a harder market
On our full public log, our model scores about 47.6% directional accuracy on China (Shanghai) across roughly 1,330 verified predictions — just below even. Here's the honest reason: it's a retail-driven, policy-sensitive market that's hard for a momentum model.
I'll start with the number I'd most like to round up. Across roughly 1,330 verified predictions on China — the Shanghai exchange (.SS) — our model has landed about 47.6% directional accuracy. That is below even (50%). It's one of our weaker markets, and rather than quietly drop it or dress it up, I'd rather show you the real figure and explain honestly why this market is harder than most. Every one of those calls sits in public at /predictions, the losing ones included, because publishing only the wins is the exact thing this company exists to refuse.
For context, the same model blends to roughly 49.7% directional accuracy across all sixteen markets we cover, and our strongest markets — Canada and the US — sit around 53%. So China isn't an outlier we're hiding; it's a market that pulls slightly below our overall line, and well under our best. China is below even. I'd rather say that plainly than bury it.
One thing to keep in mind before any number: our backtests look better than this. Backtests almost always do — they're run on history the model has already seen, with none of the messiness of predicting a future that hasn't happened yet. The live, verified record is the honest, lower number, and it's the only one I'll quote as real. On China, that real number is 47.6%.
Our edge is technical continuation, and this market gives it less to work with
Most of our features are price-and-volume momentum: the model reads how a stock has been trading and extrapolates the near-term tendency. That assumption — that a real move, once underway, keeps going long enough to be readable — is exactly what a retail-driven, policy-sensitive market makes harder to rely on. Two structural features of this market work against a continuation model.
Order flow is heavily retail. Turnover on the mainland is dominated by individual investors — a share commonly cited above 80% of daily volume, far higher than the institution-led US tape. Retail-heavy flow is noisier and more sentiment-driven on short horizons. The technical patterns my model is trained to detect are weaker signals when the marginal trader is reacting to a social-media rumour or a headline rather than to the price history the model is reading. The signal isn't absent; it's just buried in more noise than a momentum model wants.
Policy moves the market in steps the chart can't foreshadow. Top-down policy signals, regulatory campaigns, and state-backed stabilization buying move this market more than any single earnings line does, and they arrive as sharp step-changes — sentiment and policy shifts that the price history simply doesn't anticipate. A model trained to read market-driven supply and demand is, in those windows, reading a tape that a policy decision can turn on a dime. My features see the past; here, the past more often than usual fails to contain the next move.
None of this is a stock pick or a forecast about where China is headed — it's structural, a description of why a momentum model finds less traction in a market shaped this way. I'm explaining the terrain, not calling the game.
Why I'm not faking confidence to cover it
The easy move would be to bury the hit rate and keep publishing calls, or to drop China and pretend we never tried. Plenty of tools do exactly that — I wrote about the incentives in why most AI stock-picking tools are lying. The dishonest version of this business is far more profitable. I'm deliberately running the other one.
A just-below-even number tells me something real and specific: our feature set finds less traction in this regime than in our stronger markets, not that the market is unbeatable. Doing better here probably requires inputs we don't yet have — policy-event tagging, flow and sentiment signals tuned to a retail tape — rather than more momentum features stacked on a market that gives momentum less to work with. That's an honest research problem, and a different model than the one we ship today. You can read exactly how the current one is built, and what it does and doesn't look at, on our methodology page.
Here's the practical takeaway, stated plainly: on China, treat our output as weak — below even, by our own public count. We label every call Bullish, Neutral, or Bearish — never as buy/sell instructions, never with a guarantee, and never with some made-up "90% confidence" number — and on this market the confidence behind that label is low. A tool that can't tell you where it's bad isn't a research tool; it's marketing wearing a lab coat. We publish the 47.6% precisely so you can hold us to it, and if it improves as we add the right inputs, you'll watch it move on that same public log, with nothing hidden.
See the evidence for yourself — download the full resolved-prediction dataset, read the live public self-audit (hit-rate confidence intervals, live-vs-backfill split), inspect every model card, or run the research tools on your own data. No hype, just the receipts.
This article is educational content about machine learning and market structure. It is not financial advice, not a recommendation to buy or sell any China-listed or other security, and not directed at any individual's circumstances. Trading Agent is a quantitative research tool operated by WU Capital Limited (New Zealand).
