REF-1256 · Cerezuela-Espejo et al., 2021 · European Journal of Sport Science · Paper · PubMed

Is my running power reliable?

In 12 trained men, Stryd power had a test-retest CV of ≤4.3% and tracked VO2 best (r ≥ 0.91); Garmin, Polar and RunScribe running power repeated poorly. Pre-2020 devices; abstract only.

B-limited, mixed evidenceMeasured on trained athletes

Cerezuela-Espejo et al. (University of Murcia, with the University of Castilla-La Mancha) asked how repeatable five commercial running-power technologies are and how well each one's power relates to oxygen uptake.

Twelve endurance-trained men ran three submaximal protocols, each manipulating speed, body weight or slope, on a treadmill indoors and on an athletics track outdoors, on two occasions, while power was recorded at the same time by StrydApp, StrydWatch, RunScribe, GarminRP and PolarV and oxygen uptake by a metabolic cart.

Repeatability was judged by the standard error of measurement, CV and ICC; validity by a linear regression between power and oxygen uptake with the standard error of the estimate (the abstract does not say which was the predictor).

Stryd was the most repeatable technology in every environment and condition (SEM ≤ 12.5 W, CV ≤ 4.3%, ICC ≥ 0.980) and had the strongest relation to oxygen uptake (r ≥ 0.911, SEE ≤ 7.3%); Polar, Garmin and RunScribe kept some relation to oxygen uptake but their low repeatability, in the authors' words, questions their suitability.

Our reading: one small study, read from its abstract, that ranks Stryd first among the devices of its time for repeatability; it does not say that any device's watts are correct, and the absolute watts of different brands were not shown to agree.

Protocol

Who: n = 12 endurance-trained male athletes.
Design: test-retest on two occasions; each occasion indoors (treadmill) and outdoors (athletics track); three submaximal running protocols manipulating speed, body weight and slope.
Devices: StrydApp, StrydWatch, RunScribe, GarminRP and PolarV, recording simultaneously (we read GarminRP as Garmin Running Power and PolarV as Polar Vantage; the abstract does not expand them); models, mounting, firmware and the difference between the two Stryd routes are not in the abstract.
Criterion for validity: oxygen uptake by a metabolic cart; a linear regression between each technology's power and VO2, with the standard error of the estimate (direction not given in the abstract).
Repeatability: standard error of measurement, coefficient of variation, ICC, from test-retest data across environments and conditions.
Not in the abstract: speeds, slopes and loads used, how body weight was manipulated, warm-up, the repeatability and validity numbers for Garmin, Polar and RunScribe, whether absolute watts were compared between devices.
Funding and interests: not given in the abstract or the PubMed and Crossref records (full text not read).

Key Results

TechnologyRepeatability (abstract)Relation to VO2 (abstract)
Stryd (StrydApp, StrydWatch)Most repeatable in all environments and conditions: SEM ≤ 12.5 W, CV ≤ 4.3%, ICC ≥ 0.980Best: r ≥ 0.911, SEE ≤ 7.3%
Polar Vantage (PolarV)"Low repeatability""A certain relationship" with VO2
Garmin Running Power (GarminRP)"Low repeatability""A certain relationship" with VO2
RunScribe"Low repeatability""A certain relationship" with VO2
Authors' conclusion: Stryd can be considered the most recommended tool, among those analysed, for measuring running power.
Our arithmetic: a CV of up to 4.3% is a typical error of about 11 W at 250 W, the same size as the SEM ceiling of 12.5 W; two runs can then differ by up to about ±30 W (1.96 × √2 × 11 W, 95%).

Mechanisms

Measured: test-retest agreement and the strength of the power-VO2 relation differed between technologies; Stryd led on both.
Authors' interpretation: the other three keep some relation to oxygen uptake, so their main problem is repeatability.
Inference: "running power" from a foot pod or a wrist is a model output from accelerations, speed and body mass, not a measured force; a strong correlation with VO2 across speeds, slopes and loads shows that it rises with metabolic demand, not that its watts are the true mechanical power or are comparable between brands.

Practical Application

Is my running power reliable? In 12 trained men, Stryd power had a test-retest CV of ≤4.3% and tracked VO2 best (r ≥ 0.91); Garmin, Polar and RunScribe running power repeated poorly. Pre-2020 devices; abstract only.
Pace by running power only on a repeatable device, and only against itself (B-, from the abstract): on a device with a test-retest CV of about 4% or less here (Stryd), a zone set on that device can be reused on that device; the abstract gives no evidence that Garmin, Polar or RunScribe power was repeatable enough for that.
Do not compare watts across brands or with cycling watts (inference): the study ranked how well each device agreed with itself and with VO2; it did not show any device's watts to be correct, and running and cycling power are computed differently.
Watch for terrain and form: the study covered treadmill and track running with slope and load changes; trail running, steep descents and fatigue were not covered, and REF-1241's authors note that Stryd's stiffness model is not validated for steep uphill running.
Name the device: these results belong to the devices and software of before April 2020; current Stryd, Garmin, Polar and RunScribe products were not tested.

Limitations

From the abstract only. No device generations or firmware, no protocol details, no numbers for the three weaker technologies, no funding or conflict statement; the full text (paywalled; no open repository copy listed by OpenAlex) was not read.
Small, one sex. 12 endurance-trained men.
Validity by correlation. Oxygen uptake is a metabolic criterion, not a mechanical one; a high r and a low SEE show tracking, not agreement in watts.
Conditions. Submaximal running on a treadmill and a track; no trails, descents, fatigue or race efforts.
Age of the devices. Published online in April 2020; firmware and algorithms of all these products have changed since.
How this entry was checked and changed
Why this level

from the abstract; a test validation study, which the rubric places at B, with n = 12 men; held at B- because the methods are not available beyond the abstract (device generations, slopes, loads, the numbers for the four other technologies) and because "validity" is a correlation with oxygen uptake rather than agreement with a measured mechanical power