Conformal intervals, honest uncertainty.
Every IOX read ships with an uncertainty band — and that band is wide on purpose. Here’s what a conformal interval is, why it’s the honest way to show uncertainty in a fat-tailed market, and why the same width shows up on every coin.
Crypto returns aren’t bell-shaped
Most textbook uncertainty intervals assume returns follow a neat bell curve. Crypto returns emphatically don’t — they have fat tails, where extreme moves happen far more often than the bell curve predicts. Build your band on the wrong assumption and it’ll be confidently too narrow, blowing past the “impossible” outcome again and again.
Calibrate on your own mistakes
Conformal prediction sidesteps the assumption entirely. Instead of trusting a formula, it looks at how wrong the model has actually been in the past and sizes the band to match — distribution-free. The promise is concrete: an 80% band is built so that about 80% of real outcomes fall inside it, whatever shape the returns take.
And it holds where it counts: on data the model never trained on, our 80% band covered roughly 80% of outcomes.
Honest uncertainty looks unimpressive
Two things surprise people about our band, and both are features. It’s wide — because crypto really is that uncertain over a week, and a narrow band would be a lie. And it’s roughly the same width on every coin — because Ioxer ranks crowding, not the size of the return, so the band reflects general market noise, not a coin-specific confidence.
A tool that shows you a tight, custom band on each coin is usually manufacturing precision it doesn’t have. We’d rather show an honest fog than a fake laser.
Common questions
What is a conformal prediction interval?
It’s an uncertainty band built with a distribution-free method: instead of assuming returns follow a tidy bell curve (they don’t in crypto), it calibrates the band on the model’s own past errors. An 80% conformal band is designed so that about 80% of real outcomes land inside it.
Why is the uncertainty band the same width on every coin?
Because the band reflects how noisy crypto returns are in general, not a coin-specific confidence. Ioxer ranks crowding, not the size of the return — so the honest band is wide and roughly the same on every coin. A tool showing tight, coin-specific bands is usually faking precision it doesn’t have.
Does the 80% band actually cover 80%?
On data the model never trained on, our 80% band covered roughly 80% of outcomes — the coverage guarantee held out-of-sample. That’s the whole appeal of conformal methods: the guarantee is distribution-free, so it survives crypto’s fat tails.
Ioxer is research, not investment advice. IOX is a crowding read — not a price prediction, not a buy/sell signal.