Race Shape › Method
How it worksThe method, and where it comes from
Race Shape is a rules engine, not a chatbot. Every pace, watt and weekly volume is derived from a published physiological model applied to a time trial you enter. No language model decides your training load, the same inputs always produce the same plan, and every number traces back to a rule that can be pointed at.
The models, and who developed them
| Running paces | Jack Daniels' VDOT model — VO₂ at velocity, and the percentage of maximum sustainable for a given duration. Easy, marathon, threshold, interval and repetition bands all come off it. |
|---|---|
| Cycling power | Andrew Coggan's power zones, set off Functional Threshold Power. FTP from a 20-minute test (×0.95), an 8-minute test (×0.90) or a ramp (×0.75). |
| Swimming | Critical Swim Speed — Wakayoshi et al., 1992 — using the two-trial protocol over 400 m and 200 m. |
| Combining sports | Training Stress Score: hours × intensity factor² × 100. One number for running, riding and swimming together, which is the only honest way to manage a multi-sport athlete's load. |
| Periodisation | Joe Friel's base / build / peak / taper structure, with A, B and C race priorities deciding how much each race disturbs the block. |
| Altitude | Corrected for the difference between where you tested and where you train — never twice, which is the error most tools make. |
The rules it will not break
- Weekly volume rises no more than 10% — 7% if you have flagged a niggle.
- Every fourth week comes down to roughly two thirds.
- Never two hard days in a row, across every sport at once.
- The long run is capped three separate ways: by weekly share, by absolute distance, and by the time you actually said you had.
- Roughly 80% of the work sits in the easy band, measured per interval rather than per session.
What it refuses to do
A plan that will not tell you bad news is not worth having. This one says, in the plan itself:
- No time trial means no paces. It prescribes by effort and tells you why, rather than inventing numbers.
- Six weeks before a marathon, no meaningful fitness can be added — and it says so.
- A goal materially faster than your current fitness supports is not happening, and it will tell you what is realistic instead.
- Strength training does not raise your VO₂max. It is in the plan for other, real reasons.
- Stretching does not prevent injuries.
- For swimming, technique beats fitness for most amateurs — it will tell you to spend money on a coach on the pool deck rather than on more sets from us.
How it is checked
Most training-plan sites ask you to trust a name. This one is testable instead, and the tests run before anything ships:
| Named engine checks | 110 — physiology against Daniels' published tables, pace ordering, ramp caps, day spacing, determinism |
|---|---|
| Generated plans swept | 3,788, across every combination of sport, distance, training age, days available and time budget |
| Sessions asserted | 370,192 — every one checked against every rule above |
| Determinism | Identical inputs produce a byte-identical plan, and that is itself a test |
What is missing, plainly
The methods are published and the implementation is tested, but we do not currently have a qualified coach reviewing the output, and no registered dietitian has reviewed the fuelling guidance. We are looking for both. Until then, treat this as a well-built application of published models — which is what it is — and not as a substitute for coaching. If you are returning from injury or have a medical condition, see a professional first.
Common questions
Is this written by AI?
No. The plans come from a rules engine — published models plus your numbers, evaluated in order. No language model chooses your training load. The same inputs always produce the same plan, which is the opposite of how a chatbot behaves.
Who is behind the training advice?
The methods are published and named below, each attributed to the physiologist or coach who developed it. Race Shape implements them; it did not invent them. We do not yet have a qualified coach reviewing the output, and we say so here rather than implying otherwise.
Can I check that it does what it says?
To an unusual degree, yes. The engine ships with 110 named checks and a sweep that generates 3,788 different plans across every combination of sport, distance, experience and available time, asserting every rule on all 370,192 sessions. Those run before every release.
Is this medical advice?
No. It cannot assess an injury, diagnose anything, or see what a clinician would see. If something hurts while you run rather than after, stop and see a physiotherapist.