What is CI90 coverage (band calibration)?

CI90 coverage is how often the real outcome lands inside the model’s 90% confidence band. The honest target is 90% — a 90% band should contain the truth about 9 times out of 10. It measures whether the uncertainty band is the right width, not whether the forecast pointed the right way.

This is band calibration, NOT a directional hit-rate

Coverage tells you whether the width of the uncertainty band is honest — not whether the forecast pointed the right way. A model can be perfectly calibrated and still be a coin flip on direction: if it says the price will be somewhere in a wide range and the price stays in that range, coverage is fine even though the call gave you no directional edge. How often the model calls up-vs-down is a separate axis (directional hit rate). Never read good coverage as "the predictions are accurate."

What the committed backtest actually shows

On Quantustik’s committed TOP-20 backtest (the checked-in artifact behind the public calibration page, measured against the model as served, with the market and sector coupling attached), the 90% bands covered 90.63% of realized prices at 3 months, 90.91% at 6 months and 90.20% at 1 year. All three land within a point of the 90% they claim, comfortably inside the well-calibrated [85%, 95%] range. That aggregate hides wide per-ticker dispersion, and the dispersion did not improve alongside it: AVGO’s 1-year band covered only 36.5%, MSFT’s 6-month band 47.6%, and MA and V covered 55.6% and 57.1% at 3 months, while other names sit near 99%. Check the per-ticker backtest accuracy for a name you actually trade, not just the average.

Both directions are failures

Coverage well below 90% means the model is over-confident — bands too narrow, reality escapes them more often than promised. Coverage well above 90% is also a failure: the bands are so wide they are uninformative, technically "always right" but useless for a decision. Honest calibration lands near 90%, not as high as possible.

Frequently asked questions

Is CI90 coverage the same as how accurate the predictions are?

No. Coverage only tells you whether the 90% band is the right width — whether reality lands inside it about 90% of the time. A model can be perfectly calibrated on coverage and still be no better than a coin flip on which direction the price moves. Directional accuracy is a separate metric.

What coverage does the committed backtest show?

On the committed TOP-20 backtest behind the calibration page, measured against the model as served, the 90% bands covered 90.63% of realized prices at 3 months, 90.91% at 6 months and 90.20% at 1 year — all three inside the well-calibrated [85%, 95%] range and within a point of the 90% they claim. Per-ticker coverage still varies widely, from the mid-30s to nearly 100%, so the aggregate is a starting point, not a guarantee for any one name.

Is higher coverage always better?

No. Coverage far below 90% means the bands are too narrow (over-confident); coverage far above 90% means they are so wide they carry no information. The honest goal is to sit near 90%.

Learn this in the Academy

See it on a ticker

Browse all S&P 500 tickers to see this metric applied to individual companies.

Related terms

Educational research only — not investment advice.