Hang Seng (^HSI) price forecast

Hang Seng is trading at 25,859. Over the next 4 hours our simulation puts it between 25,618 and 26,260— that's the 10th-to-90th percentile of 20,000 modelled outcomes, so roughly four times in five it should land inside that band. The model puts the odds of finishing higher at 62%. These are probabilities, not predictions of certainty, and the backtest below shows exactly how often the model has been right.

Hang Seng forecast by timeframe

Intraday horizons are free. Longer horizons — 1 day through 5 years — are part of Pro.

How accurate has this model been on Hang Seng?

Every forecast site claims accuracy. Here is ours, measured by walk-forward backtest on data the model never saw during training. Where it has no edge over a naive baseline, we say so.

Backtested on 1880unseen samples. "No edge" means the model was no better than assuming the recent trend continues — an honest result, and a common one.

What the model is saying about Hang Seng

The model sees the ^HSI, currently near 25,859, drifting only modestly over the near term: about 0.38% over one week (expected 25,956) and 1.44% over one month (expected 26,231), stretching to a 19.93% gain over one year (expected 31,013). Confidence is thin. Near-term odds of an advance sit at roughly 50% for one week and just 38.5% for one month, rising to 76.9% at one year. Ranges are wide — the one-month band spans 23,985 to 28,719, and the one-year band runs from 22,616 to 42,561. Annualised volatility of 24.48% explains those broad ranges, and trailing-month momentum of 11.83% has been positive. Bluntly, the ML model shows no measurable edge: it fails to beat the naive baseline at every horizon (49.6% vs 50.4% at one day, 49.3% vs 51.0% at one week, 47.9% vs 51.6% at one month). Reliability is flagged moderate, low, and very low respectively. Given that lack of edge and the very low reliability at longer horizons, any volatility shift or momentum reversal could readily invalidate this outlook.

How to read this forecast

We don't publish a single price target, because nobody can know one. Instead the engine runs 20,000 simulated futures for Hang Seng and reports the band containing the middle 80% of them — the 10th to 90th percentile. A wide band means genuine uncertainty; a narrow one means the recent range has been calm. The odds higher figure is a calibrated probability: when we say 60%, outcomes of that kind should finish higher about 60 times in 100. Near 50% there is no directional signal at all, and the range itself is the entire forecast.

More depth: what a forecast cone is, how to read a P10–P90 range, and what makes a probability trustworthy. Our public accuracy scorecard shows where the model has an edge — and where it has none.

Other index forecasts

Forecasts are probabilistic estimates from historical price data, not financial advice. Short-horizon direction has only a small measurable edge; long-horizon numbers reflect drift plus widening uncertainty and should be read as ranges, not targets. Figures update hourly. Coneview provides statistical market information, not financial advice.