Determination of Critical Power Using a 3-min All-out Cycling Test
Participants
n=10 habitually active subjects
Protocol
Ramp test
2 × 3-min all-out tests against a fixed resistance -> end-test power (EP) and work done above EP (WEP)
5 constant-work-rate tests -> critical power (CP) and W' via the work-time and 1/time models
Key Results
| Parameter | 3-min test | Traditional CP | Agreement |
|---|---|---|---|
| End Power (EP) | 287 ± 55 W | 287 ± 56 W (CP) | not different, p = 0.37; r = 0.99 |
| WEP | 15.0 ± 4.7 kJ | 16.0 ± 3.8 kJ (W') | not different, p = 0.35; r = 0.84 |
Standard error of estimating CP from EP: approximately 6 W; in 8 of 10 subjects EP agreed with CP to within 5 W.
Power in the 3-min trial declined to a steady level within 135 s, i.e. was stable over approximately the last 45 s.
Practical Application
In 10 habitually active subjects, the end power of a 3-min all-out cycling test (287 ± 55 W) matched independently measured critical power (287 ± 56 W; p = 0.37, r = 0.99), with a standard error of about 6 W and 8 of 10 within 5 W; the work above end power (15.0 ± 4.7 kJ) matched W' (16.0 ± 3.8 kJ; r = 0.84). One 3-min all-out effort can therefore replace a multi-day series of constant-power tests for estimating CP, with W' the less precise of the two estimates.
Test conditions that the result rests on: a truly all-out start with no pacing, a fixed resistance, and reading the power of the final 30 s as end power (full text; not checked against the abstract).
Inference, not tested: the sample was habitually active, not trained cyclists — a trained athlete should expect the same logic but should confirm the EP against a known sustainable power before building zones on it.
Not from this paper; the base's own practice: a standard warm-up before the test and a re-test when training status changes.
Limitations
n=10 habitually active subjects; agreement in trained or elite cyclists is not shown here.
Two 3-min tests per subject; W' estimates (r = 0.84) are less precise than the CP estimate.
Ten habitually active participants, not trained cyclists; the end-test power = CP agreement in trained and elite riders comes from later work, not from this paper.
The test is valid only when ridden all-out from the first second; a paced attempt under-reads CP.
Agreement is at the group level (standard error ~6 W); 2 of 10 participants differed by more than 5 W.
How this entry was checked and changed
Why this level
regraded 2026-09-27 from A: test validation study, n=10
Content correction 2026-09-27
Tier "2-3 (trained cyclists)" -> abstract says habitually active subjects: Tier 1.
"Accuracy ±5 W" -> abstract: standard error about 6 W, 8 of 10 within 5 W.
Added p = 0.35 for WEP vs W' and the 135-s stabilisation from the abstract; plan items ("Week 4.5 and Week 8.5", "5 min WU") removed or marked as the base's practice.
Level A -> B (test validation study, n=10) per KB-LEVEL-RUBRIC.