I went for a 150-minute ride with Fitbit Air, Whoop and an ECG Strap. Here’s what I found

Today, I looked at three things. Firstly, I delved deeper than usual into an under-discussed stats issue in wearable reviews; then I looked at what the ECG data from Frontier ZONE actually showed during a long ride; and finally, I looked in the mirror at my haggard self after a gruelling week of training. Let’s go.

Are the Devices Accurate?

Yes. I’ve reached the point of consistency for ‘accuracy issues’ with Fitbit Air, which is generally OK but sometimes not. Today, Frontier ZONE, Whoop MG, Fitbit Air and Amazfit Helio Strap appear equally accurate for sports tracking outside of laboratory requirements, at least for steady endurance efforts of this kind. As a chest strap, Frontier ZONE benefits from direct electrical measurement rather than optical sensing, which gives it an edge during intensity changes and in general use.

For visual comparison, you can see that the Whoop MG (biceps) on this chart has one blip in the middle section, while the Fitbit Air (wrist) shows numerous minor deviations from what appears to be the consensus.

There’s a but coming.

 

 

Lies, Damned Lies and Statistics

So, the chart you’ve just seen above is a slight misrepresentation on my part. Sorry! Let me explain.

What I did was shift the Frontier ZONE’s data by +15 seconds. This makes the chart look prettier, with the curves near perfectly aligned, but it also affects more detailed analysis in dcrainmaker’s analyser (beta) tool, because the stats look at data pairs. If you move one of the time series, the paired data will differ.

In my opinion, I did the right thing, as I am trying to find what four devices thought at a given instant in time. However, by doing so, I could be masking a flaw or lag in the data, or a simple misstatement of the time (in other circumstances, perhaps the GPS time on a watch isn’t updated). What I can say is that this strongly suggests a timestamp alignment issue between datasets rather than a physiological measurement error. It is unlikely that the other three devices would all share a 15-second offset in the opposite direction, which points to the Frontier ZONE as the source of the mismatch, but I cannot be certain.

In other cases, not here, a reading might reflect a physiological delay at the points of sensing, such as the pulse transit time (typically sub-second), or there could be delays caused by processing in the sensor’s algorithms.

I know that, obviously, dcrainmaker is fully on top of the data pair issues here with his tool, and probably so is Rob (@thequantifiedscientist), but how much of the other data out there is correct? Time-series alignment is a known but rarely-discussed source of error in wearable comparisons.

That all sounds very technical. But so what?

Well, the “so what” is that the answers are different. Firstly, check out the x-axis I used. I often record data before and after an exercise and then crop down to the exercise itself, giving devices time to warm up—or, more accurately, making sure I remember to turn them all on, which can sometimes be more convoluted than you’d imagine. So the cropped data series gives different answers from the full one. Again, I’ve shown what I believe is correct.

Coming back to quantifying the timestamp adjustment. Here is an interpretation of the data before the 15-second timestamp adjustment:

• Average heart rate was virtually identical across all four devices: Frontier ZONE (120.2 bpm), Whoop MG (119.4 bpm), Fitbit Air (118.9 bpm), and Amazfit Helio Strap (118.8 bpm).
• The closest agreement was between the Amazfit Helio Strap and Fitbit Air, with a bias of just -0.1 bpm.
• Whoop MG also tracked extremely closely to the Helio Strap, with a bias of -0.6 bpm and the tightest limits of agreement (−4.1 to +2.9 bpm).
• The Frontier ZONE device recorded slightly higher values overall, averaging 1.4 bpm above the Helio Strap.
• All pairwise comparisons showed excellent agreement, with biases below 1.5 bpm.
• Overall, all four devices produced highly comparable heart-rate data during the ride, with no practically significant differences in average heart rate.

And here is the comparison after the adjustment:

• The Amazfit Helio Strap vs Frontier ZONE comparison improved from limits of agreement of −10.1 to +7.3 bpm to −4.5 to +1.6 bpm, while bias changed only slightly (−1.4 to −1.5 bpm).
• The Frontier ZONE vs Whoop MG comparison also tightened markedly, from −7.4 to +9.2 bpm to −3.0 to +4.8 bpm, with bias unchanged at +0.9 bpm.
• These much narrower limits of agreement suggest that the original differences were largely due to a timing offset rather than heart-rate measurement error. The bias barely changed; the limits of agreement narrowed dramatically — a pattern characteristic of phase misalignment rather than systematic sensor bias.
• The results for the Helio Strap, Whoop MG, and Fitbit Air are essentially unchanged, with biases of −0.6 bpm, −0.1 bpm, and +0.4 bpm, respectively.
• Following the timestamp correction, the Frontier ZONE now shows agreement much closer to that of the other devices, with all pairwise comparisons remaining in the excellent category.

Assuming the timestamp shift is the correct action, this raises the question: “How on Earth do you know what the time shift is?” There will obviously be algorithmic ways to determine the required shift, but… Jeez, good luck, Ray! I do it by visually aligning peaks and troughs, but what if those visual cues are wrong in one of the sensor’s data points?

@Ray, @Rob. What do you think?

I still think visual comparisons are often the best to make. Showing correlations can mask when the disagreements occurred, and those points (or ranges of points) can represent a specific use case the athlete was doing, e.g., running on gravel. So the real interpretation would then be “it’s running on gravel that causes XXX a problem”, not that XXX only had 95% agreement.

the ECG

The Frontier X web app gives a worrying overview. 9.5% of my heart rhythm was ‘Other’. Shocking the first time you see it on your own heart data — but ‘Other’ covers beats the algorithm cannot confidently assign to normal sinus rhythm: ectopic beats, minor arrhythmias, movement artefacts, that sort of thing. A few per cent is common and usually benign. Specific clusters at a particular point in a workout are worth discussing with a doctor, as are alerts that Fourth Frontier automatically detects. In my case, the distribution looks OK, and the events (shown below) occurred before the ride started.

new ECG metrics

A novel metric on show here is ‘strain’, which differs from how ‘strain’ is defined on other platforms. Here, it refers to heart strain when analysing the ST segment of an ECG trace.

I’ve covered ST segments here before. The ST segment is the brief pause on the ECG trace between the heart contracting and beginning to recover. During exercise, an ST-segment deviation means your heart muscle isn’t getting enough blood, which, if ignored, can be fatal. Handy to know! And that’s precisely what Fourth Frontier’s ECG straps do — more info: Fourth Frontier Zone Review (or the older X2 Review).

As you can see, my heart strain is probably OK, and the higher values are an anomaly from before the workout started. Probably.

Breathing Rate and Ventilatory Zones

Another interesting metric that pops out is the breathing rate. The ECG detects breathing by reading two related signals: as the lungs fill and empty, the chest wall moves and the ECG electrode position shifts slightly, changing the electrical signal; separately, the heart rate itself speeds up and slows down subtly in sync with each breath, a well-established phenomenon called respiratory sinus arrhythmia (now you know…I’ll test you later).

Garmin derives breathing rate using a similar method, based on HRV from a chest strap.

Fourth Frontier claims to have correlated its determination of breathing rate with an ability to define VT1 and VT2 — essentially leading to ventilatory training zones which are an excellent complement to HR and power zones over long duration, where, for example, cardiac drift impacts the usefulness of HR zones. I need to look into this some more before writing a review, as I know there is a competing and seemingly unique method put forward by Tymewear VitalPro, which combines an ECG/HRV with a strain gauge on the rear of the chest strap — the strain gauge is used to determine lung expansion, and hence an estimate of volume and breath rate can be inferred. That company also claims correlations with VT1 and VT2 (they supply Jumbo-Visma and TymeWear might also be in the news some more in the coming weeks — just saying).

At the top of the same workout page is the ECG trace as shown below. You can see where the white circular cursor is on the heart trace. I manually located it above one of the little white camera icons — these indicate whether events are automated or manual and are triggered by pressing a button on the strap. The 20-second ECG trace from that period is then zoomed in and split over two 10-second periods as shown at the top of the image. Thus, if an event occurred, you could show it to a cardiologist, who would determine whether further investigation is required. In my case, the events are unknown or noise.

Note: a reader has identified that the ECG trace in this post is inverted, i.e., I wore the strap the wrong way around. The ST strain values shown may therefore be unreliable. I am retesting and will update this section.

Other Research

Aside from what you’ve read in this article, I made a somewhat startling discovery: when compared with up to 4 devices used simultaneously, Fitbit Air nearly always records only 20% of the data points.

So it claims to record once every two seconds, but that is a maximum, and in pretty much all my tests, it’s one data point per 5 seconds. Is recording 20% of your data accurate? Opinion: That’s OK at rest, but not during sport.

Further, the 20% it records is pretty accurate, and when moved to the biceps, its accuracy was very close to that of a chest strap in one test.

In some situations, it is undeniably susceptible to different types of motion artefacts. The only bad scenario for me was on the treadmill with what appeared to be cadence-lock.

See this for more details: Fitbit Air Isn’t Truly Accurate

Take Out

Three take-outs today.

• Firstly, regular readers will be pleased to know that the ECG says I will be here to test more devices in the future. My detractors — less so.
• Secondly, statistical presentation matters more than most reviewers acknowledge. We saw that (correctly) moving a single time series by 15 seconds changed the limits of agreement from ±10 bpm to ±4.5 bpm on the same data. That is not a rounding issue — it is the difference between a device looking OK and one that looks excellent. How much published reviewer-grade accuracy data out there silently carries a similar flaw? I don’t know. Neither does anyone else.
• Thirdly, Fourth Frontier has been around for several years, and mainstream sports-tech brands have yet to produce anything comparable to its ECG metrics. A consumer-grade chest strap that monitors ST-segment shifts in real time during exercise, exports clinician-ready ECG traces, and correlates breathing mechanics with ventilatory training zones is a capability gap that Garmin, Wahoo, and Polar have not closed. I think they are working on it, but that is an educated guess. When they do, they will be following Fourth Frontier’s template.

More: TymeWear VitalPro

---

Scientific References

These are references cited by the brands mentioned here.

Fourth Frontier — cited in product and clinical documentation

• Stöggl, T. and Sperlich, B. (2014). Polarised training has a greater impact on key endurance variables than threshold, high-intensity, or high-volume training. Frontiers in Physiology, 5, 33. PMID 24550842
• Fourth Frontier Technologies / University of Rochester Medical Center (2026). ECG validation against Holter monitoring for atrial fibrillation (URMC-AF-VALIDATION-2026). Presented at ISCE 2026. Sensitivity 94.4%, Specificity 97.9%.
• Liquidia Technologies / Fourth Frontier (active). Cardiac effort and biomarker evaluation in PAH therapy. ClinicalTrials.gov NCT07285655.
• University of Strathclyde / Fourth Frontier (active). Real-world physiological monitoring in COPD. ClinicalTrials.gov NCT06419062.
• Fourth Frontier / Jessa Hospital, KU Leuven (ongoing, FX2-EXERCISE-VALIDATION). Comparative performance of wearable ECG devices during exercise in endurance athletes. Prospective observational cohort study with crossover device comparison.
• AHA abstract (2023). Real-world physiological response analysis. Circulation, 148 (Suppl. 1), 16707. DOI 10.1161/circ.148.suppl_1.16707.
• JMIR Research Protocols (2025). Cardiac effort and functional capacity in pulmonary hypertension (ERJ-00608-2023). doi:10.2196/79503.

Tymewear — cited in product and validation documentation

• Stöggl, T. and Sperlich, B. (2014). Polarized training has greater impact on key endurance variables than threshold, high intensity, or high volume training. Frontiers in Physiology, 5, 33. PMID 24550842 — cited by Tymewear in support of the 2.4x training effectiveness claim.
• Tymewear (2024). Internal validation study: Tymewear VitalPro vs Cosmed K5 metabolic cart for breathing rate and minute ventilation during incremental cycling. 26 participants. Breathing rate MAE 1.2 breaths/min; minute ventilation correlation r = 0.973. tymewear.com/pages/validation-study.
• Hicks, D. et al. (2022). Is the Tyme Wear Smart Shirt Reliable and Valid at Detecting Personalized Ventilatory Thresholds in Recreationally Active Individuals? Sensors, 22(3), 1097. PMC8835019 — independent peer-reviewed validation of an earlier Tyme Wear product for VT1/VT2 detection.

Quick answers

What is the ST segment on an ECG?
The ST segment is the brief pause on the ECG trace between the heart contracting and beginning to recover — the flat line that follows the main spike of each heartbeat. During exercise, a deviation of that flat line above or below its baseline can indicate that the heart muscle is not receiving enough blood for the workload being demanded.

Is an ST-segment deviation during exercise dangerous?
It can be. An ST deviation during exercise is the electrical signal associated with myocardial ischaemia — a shortage of blood to the heart muscle. If the cause is not addressed, a heart attack and, in some cases, cardiac arrest can follow. Most people who experience this have no prior diagnosis and no warning. That is the clinical value of real-time monitoring. If you observe persistent or symptomatic ST deviation, seek medical input — these devices are monitoring tools, not diagnostic instruments.

What does the 'Other' rhythm classification in the Fourth Frontier app mean?
The app classifies each beat of your ECG as a normal sinus rhythm or one of several other categories. “Other” covers beats that the algorithm cannot confidently assign to a normal pattern. A small proportion — typically a few per cent — is common and usually benign, often caused by ectopic beats, minor arrhythmias, or movement artefact during exercise. A persistently elevated proportion, or one that clusters at a specific point during a workout, is worth discussing with a doctor.

What is heart strain as measured by the Frontier ZONE and Frontier X2?
“Strain” on the Fourth Frontier platform refers specifically to ST-segment deviation, not to the training-load definition used by platforms such as TrainingPeaks or Whoop. The device measures how far the ST segment shifts from its baseline during exercise. A shift of 0.2 millivolts or above is the threshold associated with potential oxygen deprivation to the heart muscle in exercise stress test interpretation. The device can be configured to alert at 0.3 millivolts, providing a margin above background noise.

How does the Frontier ZONE derive breathing rate from an ECG?
Two mechanisms are at work. First, as the lungs fill and empty with each breath, the chest wall moves and the electrode position shifts slightly, producing a subtle change in the ECG signal that can be extracted algorithmically. Second, the heart rate itself speeds up and slows down slightly in sync with each breath — a well-established phenomenon called respiratory sinus arrhythmia. Both signals encode breathing frequency, and the device uses them to estimate breath rate without a separate respiratory sensor.

What is the difference between breathing-rate zones and heart rate zones for training?
Heart rate zones are influenced by hydration, heat, fatigue, and caffeine, and drift upward over long sessions even at a stable workload — a phenomenon known as cardiac drift. Ventilatory thresholds (VT1 and VT2), derived from breathing mechanics, reflect the body’s actual metabolic crossover points more directly and are more stable across varying conditions. For long-duration training in particular, breathing-rate zones can provide a more reliable intensity signal than heart rate.

How does the Tymewear VitalPro measure breathing differently from the Frontier ZONE?
The VitalPro uses a strain gauge on the rear of the chest strap to detect the physical expansion of the rib cage with each breath, allowing it to estimate not just breathing rate but also tidal volume and minute ventilation — the total volume of air moved per minute. The Frontier ZONE derives breathing rate from the ECG signal alone. Both approaches claim correlations with VT1 and VT2, but the underlying sensor technology and the richness of the respiratory data they produce are different.

Why does a 15-second timestamp offset matter for heart rate comparison accuracy?
Accuracy tools such as the DCRainmaker Analyzer calculate statistics by pairing data points at matching timestamps from two devices. If one device’s timestamps are offset by 15 seconds, the paired points no longer represent the same instant in time. During sections of a ride where heart rate is changing rapidly, this produces large apparent disagreements that disappear entirely once the offset is corrected. The bias barely changes; the limits of agreement narrow dramatically. It is a statistical artefact caused by phase misalignment, not a measurement error.

Can I share Fourth Frontier ECG data with my cardiologist?
Yes. The Frontier X2 and Frontier ZONE allow export of up to three hours of continuous ECG trace as a PDF from the online dashboard. A cardiologist can review this in the same way they would any single-lead ECG recording. The devices are not replacements for clinical investigation, but the exported trace — particularly if recorded during a hard effort when symptoms were present — is a meaningful supplement to a clinical conversation.

Is the Frontier X2 or Frontier ZONE an FDA-approved medical device?
No. The Frontier X2 and Frontier ZONE are classified as wellness and fitness devices, not medical-grade diagnostic tools. The separate Frontier X Plus product from Fourth Frontier holds FDA 510(k) clearance as a prescription-based medical ECG monitor for arrhythmia detection. The X2 and ZONE are not intended or certified for clinical diagnosis.

This article is part of the site’s sports science reference for endurance athletes.

For the full Amazfit range, accuracy data and Zepp Health analysis: Amazfit sports watch guide. For full context on the Fitbit Air within the Google ecosystem: Google Fitbit and Wear OS hub.

Last Updated on 10 July 2026 by the5krunner