Most bike power meters read within a few percent of true power and are highly consistent, but individual units can be more than 5% out and single-sided meters add error from left-right asymmetry
Interpretation
The largest test (Maier 2017, 54 meters from 9 manufacturers) found mean trueness of -0.9% ± 3.2% with 6 meters beyond 5% and high precision (CV 1.2%). Smaller comparisons against a reference crank meter (Bouillod 2017, one male cyclist; Nimmerichter 2017, 6 cyclists) found good steady-state agreement but larger errors in sprints, standing and vibration. Single-sided meters were valid on average (Granier 2020, 11 trained subjects; Valenzuela 2022, 33 male cyclists) but their error tracked pedalling asymmetry of 4-6%, rising to 10-13% at 100 W. Graded moderate: consistent direction, but small samples, trained (mostly male) cyclists, and models from 2017-2022 that have since changed. No trial has tested whether training with power improves age-group triathlon results.
Practical implication
A power meter is useful for intervals, hilly courses and race pacing, but it doesn't build fitness by itself. Zero-offset it, check it against a second source such as a direct-drive trainer, keep zones on the same meter, and treat single-sided numbers as estimates. Findings come from trained cyclists, not age-group triathletes, and specific models date quickly.
Sources
6 sources, independently verified.
Against a first-principles treadmill cycling model, calibrated power meters showed a mean trueness of -0.9% ± 3.2%, with 6 of 54 deviating by more than 5%; precision (coefficient of variation) was 1.2% ± 0.9% and differed between manufacturers. Authors advise checking each meter against a first-principles reference. Models of the time (2017).
Hub, single-sided crank and pedal meters were as reproducible as a reference crank meter, but the pedal meter read 36.9% low in 5-s sprints in a low gear and 4.5% low in the heavy zone; the single-sided crank meter read 5.1% and 4.9% low in heavy and low-severe zones and 4.4% lower standing than seated; both read lower under strong vibration. Single rider; not generalisable.
Dual-sided pedal meter vs reference crank meter at 100-300 W, a 5-min time trial and a 30-s field sprint: no significant difference in mean power, limits of agreement about 1.0 ± 5.0 W, typical error 2.9% for steady power and 7.4% in 30-s sprints; CV about 3% vs 1% for the reference. Very small sample.
Left-crank single-sided meter vs reference crank meter from 100 to 1,250 W: trivial to small differences; difference correlated with pedalling asymmetry (r = 0.58), and asymmetry correction helped only marginally; CV 2.1% vs 1.3% at submaximal intensities. Suitable unless a typical error under 3% is required.
Left-only pedal meter vs dual-sided version across 100-500 W, sprints, three cadences, seated and standing: no significant differences, ICC at least 0.90, CV mostly 5% or less (5-7% at 100 W). Left-right asymmetry 4-6% in most conditions, rising to 10-13% at 100 W. Valid on average; caution for individuals with high asymmetry. Men only.
Mobile power meters provide valid, reliable field measurements and are useful for monitoring training and race demands, but simple summary methods cannot reflect a session's variable power distribution or its metabolic stress, and current methods of quantifying training load are inadequate.