Your race, taken apart.

Most running apps give you a score. Iteros gives you a diagnosis: what actually went wrong, what did not, and what it could not tell either way.

What one race file gets you

12 failure modes, each one tested
Pacing, fuelling, heat, terrain, the taper, the start — every mode gets its own number and its own evidence, not one headline cause with the others denied in prose.
Your data, modelled against a repository of more than a hundred thousand runners' races and training blocks
140,000+ across three sources: race results from 5K to the marathon, marathon blocks with the training behind them, and one marathon field’s five-kilometre splits. The thresholds come from the literature rather than from a hunch. Execution scoring, the pacing brackets and the wall comparison are marathon-only — that is the one set with splits, 47,998 finishers, and nothing equivalent exists at another distance yet.
It names what it could not test
A mode with no evidence behind it is reported as untested, with the thing that would settle it — never quietly scored zero and averaged in. That is the part most of this is built around.

It does not claim an accuracy figure.: Sensitivity was measured once, on a cohort whose own labels proved unreliable. Until that is redone on labels worth trusting there is no number here worth quoting.
The race prediction is not calibrated.: It is a model built from your own training, and no predicted result has yet been frozen and checked against what happened. When that exists it will be shown; it does not exist now.
The distances are not equally well served.: The marathon is the best supported. Two of the 12 modes rest on marathon split data and cannot be ranked at shorter distances at all, and the analysis says so on the page rather than filling the gap.

Everything is measured from files you already have — Garmin, Coros or Strava exports. Create an account to keep a history, or sign in.