REF-040 · Morgan et al., 2018 · Journal of Sports Sciences, Vol. · Paper

Road Cycling Performance and Critical Power Predictive Validity

Blimited evidenceMeasured on trained to highly trained athletesWritten before October 2026, being re-checked

Participants

n=12 competitive male cyclists. Each completed an incremental test to exhaustion, an FTP test, 4-5 self-paced time-trial (TT) prediction bouts on a stationary bike in the lab, and a 16.1-km road TT, all with mobile power meters.

Protocol

Validation study: twelve competitive male cyclists (from the abstract).
In the lab: an incremental test to exhaustion, an FTP test and 4-5 self-paced TT prediction bouts on a stationary bike; CP and W' taken from the power-duration relationship.
In the field: a 16.1-km road TT with mobile power meters; outcomes were actual vs predicted completion time, and the agreement between FTP and CP (limits of agreement).

Overview

Morgan et al. asked two questions: can CP and W' derived from self-paced TT prediction trials on mobile power meters predict a 16.1-km road TT, and do FTP and CP agree? The predicted completion time (27.5 ± 3.3 min) was not significantly different from, and correlated with, the actual road TT time (26.7 ± 2.2 min; r = 0.89, P < 0.01).

CP and FTP were not significantly different as group means (275 ± 40 W vs 278 ± 42 W, P > 0.05), but the individual limits of agreement were +30 to −36 W — too wide, in the authors' words, to consider FTP and CP interchangeable.

Key Findings

Predicted vs actual 16.1-km TT time: 27.5 ± 3.3 vs 26.7 ± 2.2 min, r = 0.89 (P < 0.01), no significant difference
CP vs FTP group means: 275 ± 40 vs 278 ± 42 W (not different)
CP vs FTP limits of agreement: +30 to −36 W — individuals can sit far from the group mean
Authors' conclusion: CP and W' from self-paced trials on mobile power meters accurately predict 16.1-km performance; FTP and CP are not interchangeable

Practical Application

In 12 competitive male cyclists, CP and W' from self-paced time-trial prediction bouts predicted a 16.1-km road TT time to within a non-significant 0.8 min (r = 0.89); CP and FTP were equal on average (275 vs 278 W) but disagreed by up to +30/−36 W in individuals — so an individual FTP is not a stand-in for CP, and a pacing model built on CP/W' should use a CP that was actually measured.
Not from this paper; the base's own practice: when a rider's FTP and CP differ by more than a few percent, look at how each was obtained (pacing of the test, anaerobic contribution, protocol compliance) before deciding which number to prescribe from, and base race-pacing models on CP/W' rather than a 20-min FTP alone.
Inference, not tested: the wide limits of agreement mean the CP-FTP gap is an individual trait; measuring both once tells you which way your own gap runs.

Limitations

n=12 competitive male cyclists; one road distance (16.1 km); no women.
Prediction accuracy is a group-level statement; the abstract reports no individual prediction error.
How this entry was checked and changed
Why this level

regraded 2026-09-27 from B+: test-validation study, n=12

Content correction 2026-09-27
Removed the invented "CP-based pacing: 92% of predicted power achieved in races", "CP-based pacing proved more reliable than FTP-based pacing" and "individuals with larger differences show pacing variability" — none is in the abstract.
Added the abstract's actual numbers (TT 26.7 ± 2.2 vs predicted 27.5 ± 3.3 min; CP 275 ± 40 vs FTP 278 ± 42 W; LoA +30/−36 W).
Testing advice (30-min MLSS / 3-min all-out / ±20 W rule) removed or relabelled as the base's own practice. Level B+ -> B (validation study, n=12).
2026-09-28: added Protocol from the abstract (Europe PMC PMID 30387374).