September 2026
Race-day plans build a fueling and pacing plan from an athlete’s own rides, with the range and the evidence grade on every hourly target. Evidence grades put that grade next to every number in the platform. Together they are the first time RAMTT shows its work everywhere at once.
Most training platforms hand an athlete a score. RAMTT has always tried to hand over a model instead, with the uncertainty carried along, and today that model reaches race morning. A plan is not a template with your weight filled in. It is built from your thresholds as they sit today, the fuel you have shown you can take in, and how much the last hour has cost you before, and every one of those inputs carries its grade.1
The plan is available to every athlete on a paid plan from today, and coaches see it in the coach workspace with the model’s answer and their own overrides side by side. The model name in the API is ramtt-plan-1.
A plan rests on four estimates the platform already keeps: fuel, thresholds, recovery and durability. Each is a range rather than a point, and the plan inherits the width. The panels below show how three of them behave as rides come in; switch between them to see the shape of the evidence.
Sample numbers to show the component. The real validation note lists the cohort, the method and the range.
Sample numbers. A carried-forward test is exact on day one and drifts; the estimate starts wide and narrows.
Sample numbers. The validation note reports the full distribution, not a single line.
The fueling target is the best-graded of the four. The literature on how much carbohydrate an athlete can take in per hour is consistent across populations, which is why the core of the fueling model carries an established gradeestablished. What it cannot tell you is what your gut will take today, and that is where the log takes over from the paper: the error against what athletes actually took in falls from about twenty grams per hour to five as rides accumulate.
Thresholds move the other way. A test is exact on the day it is taken and drifts every day after; an estimate from ordinary rides starts wide and narrows. After three weeks the estimate is usually the better number, and the plan says which it is usingsupported.
We checked the plan against forty riders over a season, and against two baselines: a fixed rule of the kind most platforms ship, and a population average. The table shows the three side by side on the four measures that matter for a plan. The RAMTT column is pinned; the marker shows the better number in each row.
| Ramtt plan | Fixed rule | Population average | |
|---|---|---|---|
| Fueling targetError against the log, g/h | 5.1after 60 rides | 23.6constant | 14.2by body mass |
| ThresholdError against a test, W | 7after 60 days | 26test carried forward | 31by category |
| DurabilityCorrelation with power fade | 0.72six long rides | 0.45heart-rate drift | 0.18by category |
| Plan followedShare of hourly targets hit, ±10 % | 81 %n = 40 | 54 %n = 40 | 49 %n = 40 |
Sample numbers to show the component. The validation note2 lists the cohort, the method and the ranges; nothing here should be read as a result.
Twelve coaches ran the plan with their athletes through the summer. Three of them are quoted below; the rest are in the note.
1 of 3
Race-day plans and evidence grades are live for every athlete on a paid plan and in every coach workspace. Plans are built for road, gravel and triathlon bike legs today; running follows when the durability engine has been validated on runners, and not beforeRamtt's own.
1 Inputs: a plan uses the fresh threshold, the within-ride drift, the fuel estimate and the durability index. Each carries its own grade and range; the plan never shows a single grade for the whole. ↩
2 Validation: forty riders, one season, plans compared against logged intake and measured power. The note reports the distribution for each measure and the riders where the plan did worse than the rule. ↩
1. The validation note for race-day plans, with cohort and method. View note
2. How evidence is graded, engine by engine. Read the document
3. The science behind fueling, thresholds, recovery and durability. Open the science page
Q.Can a coach change the plan?Yes. An override is shown to the athlete with the coach’s name, next to the model’s own answer. Neither is hidden.
Q.What happens with no fueling log?The plan uses the population average, says so, grades it accordingly, and asks for a log after the first ride.
Q.Does the hard stop before a race ever lift?No. It is one of the few rules we keep, and it lifts the moment the race is logged, not before.