Make uncertainty
useful.
Explore future price distributions, inspect the assumptions behind each projection, and compare forecasts against outcomes that were unseen at their historical origins.
Estimate.
Test.
Disclose.
Model ranges are not calibrated. Forecasts are neither guaranteed price targets nor predictions of the next change-point date.
The projection room.
Price paths / probability fan
Preparing forward model
Reading completed weekly history and fitting the exploratory model.
Inspect a future week.
Pick a projected week to highlight its position on the chart and read its model-conditional price range.
Load a valid historical series to inspect its forward projections.
Show your working.
Test windows overlap at horizons above one week. Commercial data rights, uncertainty calibration, independent historical samples and real-world predictive usefulness remain unverified.
Proof before confidence.
We will need immutable lawful datasets, multiple regimes, benchmark comparisons, and properly calibrated uncertainty before describing these forecasts as reliable.