Price error
Absolute percentage error measures the distance between a central forecast price and the first stored market reference on or after its target date.
Oil forecast track record.
An auditable record of matured model scenarios. It reports errors and misses as well as correct directions; it is not a claim of predictive certainty.
Aggregate matured performance.
Matured scenarios scored
3
Mean absolute % error
9.22%
Direction correct
66.7%
Inside stated range
0.0%
Outcomes available through 02 Nov 2026. Metrics use the newest stored forecast for each benchmark, horizon and target date.
Recent matured forecasts.
| Target | Market | Horizon | Forecast | Outcome | Absolute error | Direction | Range |
|---|---|---|---|---|---|---|---|
| 2026-11-02 | WTI | 5 sessions | $108.77 | $97.30 | 11.79% | Missed | Outside |
| 2026-10-30 | WTI | 5 sessions | $111.76 | $104.12 | 7.34% | Correct | Outside |
| 2026-10-30 | Brent | 5 sessions | $127.73 | $117.70 | 8.52% | Correct | Outside |
What the scorecard can and cannot show.
It compares the central scenario with the first stored market observation on or after the target date. Absolute percentage error measures distance, direction asks whether both moves had the same sign, and range coverage checks whether the outcome fell between the published low and high. None proves that the model will work next time.
Inspect the current oil outlook, the latest evidence brief and our methodology.
Three ways the scorecard checks the model.
Direction
Direction is correct only when the predicted move from the reference and the later observed move have the same sign. It does not reward a close price with the wrong direction.
Scenario range
Range coverage records whether the later value was inside the published low-to-high interval. A wide range may cover more outcomes without making its centre more accurate.
How to read the record over time.
Do not judge the model from one correct call or one miss. Compare a meaningful run of matured observations, separate WTI from Brent, and compare short and long horizons because each can behave differently. Direction accuracy should be read beside price error: the model can predict the right direction but still be far from the observed price.
Range coverage also needs context. A very wide interval is easier to hit, so coverage is not useful by itself. The most informative record combines a reasonably narrow range, a central estimate with low error and transparent evidence available before the target date. New scenarios remain excluded until their target date has passed and a later stored observation exists.