RAMTT now grades every number it shows you
From today, every estimate in RAMTT carries a grade next to it. Not a confidence score, and not a colour, but a word that says what the research behind the number looks like: established, supported, emerging, early, or our own. The grade sits where the number sits, in the app, in the weekly review and in the race-day plan, and it changes when the evidence changes.
We have written before about why we prefer judgment over rules, and about the difference between a count and a model. This is the piece of that argument we had not shipped: the part that lets you check us. A platform that asks you to trust its judgment owes you the reasoning, and the reasoning starts with knowing what each number stands on.
What the grade means
The grade describes the state of the field, not our confidence in you. An established claim can still be wrong for a particular athlete; that is what the range around the number is for. The grade tells you how much the field knows about the mechanism, and the range tells you how much we know about your case. The two are shown together and they answer different questions.
- Established. Replicated across many studies and populations, with mechanisms that are understood.
- Supported. Consistent findings, a plausible mechanism, and open questions about how far it generalises.
- Emerging. Promising, with results not yet replicated or that depend on a specific population.
- Early. A small evidence base or a single line of work. Useful, provisional, and marked as such.
- RAMTT's own. Models we built and are validating against our own athletes, not yet published.
We grade at the level of the claim, not the feature. One screen can carry three numbers with three grades, and it usually does. The threshold on a ride might be established; the estimate of how far it drifted in the last hour might be supported; the reserve at the top of the climb might be our own. Showing one grade for the whole screen would mean choosing the best or the worst, and both would mislead.
How a grade is set
Two people who did not write the engine set the grade against a written rubric, and the disagreement between them is recorded. A grade the two do not agree on is set to the lower of the two until it is resolved. A replication that confirms a finding can raise a grade one step and never two, because the next step needs a different population. A failed replication lowers it immediately.1
What the platform does while a claim is early
An early claim is shown, not hidden. It carries its grade, it carries a wider range than it otherwise would, and it is never the only reason for a recommendation. If the only evidence for a piece of advice is early, the advice is phrased as a question rather than an instruction, and the app tells you what would settle it. A coach can turn early claims off for an athlete entirely, and some do.
The grade is the first thing I look at now. Not because I distrust the number, but because it tells me how hard to argue with it.
Coach, road cycling, Denmark
What changes for you
Nothing in the model has changed; what has changed is that you can see it. If you have been following a fueling target you will find it graded established, with the ladder that raises it graded emerging. If you have been reading the recovery balance you will find it graded low, which is where it belongs until our own validation is done. And if you find a grade you disagree with, the science note for the claim has the studies and the rubric, and a way to tell us.
This is the beginning of a longer piece of work. Over the autumn the grades will start to move as replications land, and each move will be written down where you can see it.
- The rubric and the grader disagreement log are published with each science note. Grades for RAMTT's own models cannot rise above supported, because our validation data is not public and cannot be checked by anyone else. ↩