Forecasting notes

Meta Prophet Quantiles: how to read forecast bands

Published February 10, 2026

Quantile bands are a discipline tool. They force you to look at uncertainty, not just a single predicted line. This note explains how to pick quantiles, how to interpret bands, and how to compare Meta Prophet against IBM TimeMixer inside the terminal.

Quick defaults

Start with `0.1,0.5,0.9`
Add `0.25,0.75` for an inner band
Compare engines, not vibes

What the bands mean

A quantile forecast answers: “at each timestamp, what price would we expect to be below X% of the time?” It is not a promise. It is a calibrated range you can compare against your process.

When the outer band widens, uncertainty is expanding. When it tightens, the model is more confident given recent dynamics. The job is to combine that information with context: trend signals, catalysts, liquidity, and risk constraints.

How to choose quantiles

`0.1,0.5,0.9` for a wide risk envelope
`0.25,0.5,0.75` for tighter planning
Use fewer quantiles when you want clarity

Model comparison

Quantura supports two engines:

  • Meta Prophet: best when trend and seasonality are meaningful signals.
  • IBM TimeMixer: useful for fast, foundation-model style baselines to compare against.

Next steps in the terminal

After you generate a run, overlay indicators and pull headlines before you make a call:

Indicators: overlay SMA/EMA and confirm momentum
News: sanity-check recent catalysts
Options: map strikes to your bands for risk planning
Learn more

The terminal is designed to reduce context switching. Pick a ticker once, then iterate: forecast, overlay, verify, record.