Copper Futures (HG=F) price forecast

Copper Futures is trading at 6.37. Over the next 4 hours our simulation puts it between 6.33 and 6.41— 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 odds of finishing higher are near a coin toss, so the model claims no directional signal here — the range is the whole forecast. These are probabilities, not predictions of certainty, and the backtest below shows exactly how often the model has been right.

Copper Futures 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 Copper Futures?

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 1835unseen 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 Copper Futures

The model projects copper (HG=F) at roughly 6.39 over one week (a 0.44% move, prob_up 52.7%), 6.47 over one month (1.66%, prob_up 61.7%), and 7.86 over one year (23.48%, prob_up 73.4%). Confidence weakens sharply with time: the one-week read is "moderate," one-month "low," and one-year "very low." The one-year band alone spans about 5.10 to 12.12, so treat the longer projections as very wide. Annualised volatility sits at 33.5%, which explains those broadening ranges, and trailing-month change is a modest +3.22%. Crucially, the ML model beats its naive baseline at none of the tested horizons — accuracy of 48.9% vs 51.1% at one day, 49.2% vs 54.1% at seven days, and 43.5% vs 57.1% at thirty days. Bluntly, the model shows no measurable predictive edge here. A volatility spike beyond the assumed 33.5%, a break in the recent momentum, or the wide distribution simply resolving toward its lower bound would all undercut 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 Copper Futures 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 commodity 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.