Fitbit Air vs Whoop MG in HYROX Accuracy Test

I don’t drink much alcohol, but tonight was going to be one of those nights. The plan? 2 hours in the gym doing something pretty hard, then hoping that the much higher excess post-exercise oxygen consumption (EPOC) will burn off the alcohol. It doesn’t. The liver clears alcohol at a fixed rate regardless of what you just did in the gym. All I did was arrive at the pub tired. Very tired.

I was very tired because I ran to and from the gym, and whilst there, did a Hyrox simulation with 5 minutes per station, plus some faffing around and chatting. Two hours from start to finish.

Other than having a few beers, the plan was to test out Fitbit Air, Amazfit Cheetah 2 Ultra, the Frontier Zone ECG chest strap, Whoop MG (biceps) and my trusty Polar SENSE (biceps).

What I did

Hyrox intersperses 8 functional exercises with 1km runs. I did my best to match the real-world arena events with the functional exercises. I had no sled to push or pull this time, so I improvised, and I might have to admit that the wall-ball target I chose was lower than it should have been, explaining why it felt much easier than normal.

What I found

Fitbit AIR was the worst of the devices on test, but I’d say it was mostly good. You’ll see why once we get past the stats and onto the charts. Here are the summary accuracy stats for each device

• Fitbit Air – Accurate average HR (within ~1 bpm of others) but widest error range (up to ±22 bpm); least consistent device tested.
• Amazfit Cheetah 2 Ultra (Zepp) – Excellent accuracy; −0.1 bpm bias vs WHOOP, +1.1 bpm vs Polar SENSE; slightly more consistent than WHOOP.
• Polar SENSE – Best optical sensor; 5,501 samples; ~+1 bpm bias vs Zepp/WHOOP; tightest agreement among wearables.
• Frontier Zone (chest strap) – Reference device; best overall agreement, but only 275 matched samples.
• WHOOP MG – Excellent accuracy; ~+1.2 bpm bias vs Polar SENSE; essentially tied with Zepp.

And here is a timeseries chart and Levels of Agreement between each pair of devices.

The highlighted part of the workout skips the run to and from the gym, so it goes straight into the ski erg and the treadmill. Clearly, several devices struggle a bit to get up to speed, even after 20 minutes of easy running. Whoop and Zepp/Amazfit both over-read during the ski erg, but the interesting observation is that Zepp toggles between the correct reading and the incorrect Whoop reading. Oddly, two incorrect values match, no? My explanation is that both have the same motion-artefact detection bug.

After that, all the devices pretty much agree, except Air, which has a slight blip on the sled pull but then undercooks things slightly on the burpees and wall balls.

I’ll keep coming back to Whoop’s accuracy. It is more than accurate enough on the biceps. I’ve found that in my review of the Whoop MG version immediately after launch and consistently since. People who say ‘Whoop is inaccurate’ mean ‘inaccurate on the wrist’ (I agree, it can be), and every device has a significantly higher chance of being inaccurate on the wrist, depending on many factors, including motion artefacts. I find Garmin to be inaccurate on the wrist – I simply rely on a chest strap instead, as I want accurate data in Garmin Connect.

The final data issue to nail is the number of data points, the ‘n’ value. For some reason, this particular non-GPS recording prevented Frontier Zone from exporting its usual FIT file, instead, the TCX (CSV) file I got and converted had only one data point every 40 seconds or so. Very odd. Very annoying. The Fitbit Air’s earlier paucity of test data points was mostly rectified here, with it now recording half as many as before, up from the 20% found in previous tests. The 50% is correct behaviour as Air is specified as up to one recording per 2 seconds (0.5Hz)

Sleep Quality – Post Exercise+Alcohol

Alcohol negatively affects sleep quality.

This particular post-beer night, I was blessed with good sleep, and even my HRV4Training HRV readings were above my normal range.

Fitbit Air did a good job of recording my sleep parameters, even catching the 6 am toilet visit. However, it rated my sleep as 96, whereas Eight Sleep came in slightly more conservatively at 86. These scores are meaningless scientifically, but considering alcohol was involved and my HR at rest throughout the night declined to a still high level by morning, I think Fitbit’s 96 is probably over-egging reality.

Take Out

The main takeaway is not to waste your time doing workouts when there is alcohol involved before or after. That aside, let’s get back to the tech on test.

Amazfit Cheetah is positioning itself as the HYROX watch. So the HR performance here is good but not perfect. Hopefully, the brand has some more tweaks for us that I will get to experience firsthand when I get hold of the new Balance (next week).

Fitbit Air keeps proving itself generally good when it actually records data, but annoying when it just forgets to – its performance seems unnecessarily overhyped by almost the entirety of the internet. Either way, it’s not an athletic-grade product. Fitness grade? Sure. Recovery grade? Sure. General tracker? Sure. Just not for athletes doing athletic stuff.

Last Updated on 8 June 2026 by the5krunner

the5krunner

tfk is the founder and author of the5krunner, an independent endurance sports technology publication. With 20 years of hands-on testing of GPS watches and wearables, and competing in triathlons at an international age-group level, tfk provides in-depth expert analysis of fitness technology for serious athletes and endurance sport competitors. ID