Copper price · World · 3 months ahead

2026-10: the model assigns an 80 percent probability to 13,040 to 15,680 US dollars per metric ton

Point estimate 13,725 US dollars per metric ton from Drift, against 13,543 US dollars per metric ton observed in 2026-07. drift beats last_value on out-of-sample MAE by 7.0 percent over 45 cutoffs (minimum 24, margin 5 percent).

ModeledMetric page

Distribution over the history

Fan chart

Solid: observed (US dollars per metric ton). Dashed and shaded: model point estimates and 80 percent intervals from empirical residual quantiles. Filled dot: the record on this page. Green or red dot: the realised value, inside or outside its interval.

Quantiles: p5 12,762 · p10 13,040 · p25 13,651 · p50 13,869 · p75 14,700 · p90 15,680 · p95 16,284
Source: empirical quantiles of 36 in-window 3-step residuals, in log space; no normality assumed

The record

Immutable fields

Key: copper-price:WLD:3
Id: 893c2ca0-3a7a-4965-a014-7b261c850976
Created: 2026-09-07 22:16 UTC
Data cutoff: 2026-09-07 22:16 UTC
Origin: 2026-07 = 13,543 (obs 156369)
Target period: 2026-10 (92 days)
Point: 13,725 US dollars per metric ton
Interval: 13,040 to 15,680 at 80 percent
Model: drift models_v1.targets_v1
Baseline: last_value
Snapshot: 82e3d11a554b5a2476e4951e747d0cfa6138e4eb3187a7f9478ffb74019594e4
Data mode: ingested_at
Status: active

Drivers

Statistical association, not causal

Drift uses only the target's own history; there are no driver contributions.

Outcome

Not yet realised

The observation for 2026-10 has not been ingested. The weekly forecast.evaluate job attaches it when it arrives; the record above does not change.

Scenario assumptions

What the forecast takes for granted

No structural break in copper-price between the data cutoff and the target period.
The source keeps publishing the series on the same definition; a rebasing or redefinition invalidates the comparison.
The average historical change per period continues.

Falsifiers

What would show this forecast wrong

If the first published value of copper-price (WLD) for 2026-10 is below 13,040 US dollars per metric ton or above 15,680 US dollars per metric ton, the 80 percent interval is falsified.
The model assigns a 10 percent probability to a value below 13,040 US dollars per metric ton and a 10 percent probability to a value above 15,680 US dollars per metric ton; interval misses should occur about one time in five over many forecasts, and a run of misses well above that rate falsifies the calibration.
The point forecast is above the last observed value (13,543 US dollars per metric ton); a realised value at or below that level falsifies the direction call.

Competing models

Scorecard for this target and horizon

Model	Status	Cutoffs	MAE	RMSE	Direction	80% coverage	Skill vs last value
Drift models_v1.targets_v1	champion	45	566.1	779.7	71%	84%	7.0%
Holt (exponential smoothing with additive trend) models_v1.targets_v1	challenger	45	785.3	999.1	49%	87%	-29.0%
Last value models_v1.targets_v1	baseline	45	608.7	825.8	0%	84%	0.0%
Linear trend models_v1.targets_v1	challenger	45	1,260	1,511	42%	87%	-107.0%
Ridge regression on lags and drivers models_v1.targets_v1	challenger	45	620.5	838.9	29%	84%	-1.9%
Seasonal naive models_v1.targets_v1	challenger	45	1,315	1,741	51%	76%	-116.1%

Model card

Drift

Purpose. Point and 80 percent interval forecasts of copper-price (WLD) 1, 3, 6, 12 months ahead. Owner: Forecast Lab (HV 3.0).

Method

The forecast extends the mean first difference of the whole training window from the last value: last value plus horizon times the average change. The series is modelled in natural logarithms and transformed back, so intervals are asymmetric in level terms.

Uncertainty

Intervals are empirical: the model is refitted at earlier origins inside the training window, the realised errors at the same horizon are collected, and the interval is the 10th to 90th percentile of those errors around the point forecast. No normality is assumed.

Training and evaluation

Training window: 415 periods, 1992-01-01 to 2026-07-01. Evaluation: rolling-origin backtest over 45 cutoffs using pseudo real-time data (the earliest ingested value per period; no vintage archive exists for this series, so revisions between the cutoff and today are not reproduced).

Known limitations

Structural breaks after the cutoff are not modelled.
Driver contributions are statistical associations, not causal effects.
Promotion to champion requires beating the last-value baseline on out-of-sample MAE by at least 5 percent over the minimum number of cutoffs; otherwise the baseline stays champion.
Retraining: every scheduled forecast.run refits on the data then available; backtests re-run from /admin/forecasts.

Version chain

Supersedes

This is the first record for its key.