Apple Watch Heart Rate During Strength Training: Why Your Numbers Look Wrong

You finish a hard set of squats, look down, and your watch says 118 bpm. You feel like your heart is trying to leave your chest. Then you sit on the bench for four minutes and it drifts up to 132 while you are doing nothing at all.

That experience is why so many people conclude their watch is broken. It usually is not broken. It is doing exactly what an optical sensor on your wrist does under load, and once you understand the failure modes, you can still get useful data out of a lifting session.

This is about what is real, what is noise, and which number to actually look at.

TL;DR: Wrist optical heart rate works better during lifting than the internet claims, but it has predictable weak spots. A 2026 study of 62 men doing resistance exercise found all four smartwatches tested (Apple, Samsung, Fitbit, Garmin) delivered usable heart rate (correlations 0.64-0.97 with ECG, ICC above 0.94, limits of agreement around ±10 bpm), and only the Apple Watch showed no statistically significant difference from ECG during resistance work. Apple's own September 2026 accuracy study deliberately included loaded forearm grip and strength training as a hard test, and reported Apple Watch beat every comparison device there. But there is a real caveat running through all of it: lag. Optical sensors smooth and average, so during short sets that jump from 90 to 160 bpm in 20 seconds, the reading arrives late and low. Calorie burn during lifting is a different story and is genuinely unreliable (one study found correlations of just 0.10-0.34 with indirect calorimetry). The practical move: do not trust the mid-set number, do trust the reading you take 30-60 seconds after the set ends, and use your overnight heart rate and HRV as the real recovery signal.

YouTube: Related video

Why lifting is the hardest test for a wrist sensor

A chest strap measures the electrical signal of your heart. A wrist watch shines green light into your skin and watches how the blood volume under the surface pulses. That second method, called photoplethysmography, depends on a few things being true: the sensor is sitting flat against skin, the tissue underneath is not moving, and blood flow near the surface is steady enough to detect a pulse wave.

Now think about a heavy set of rows. Your forearm muscles are contracting hard, which squeezes the blood vessels and changes local blood flow right under the sensor. Tendons in your wrist move under the band. Your wrist flexes and extends with every rep. And your actual heart rate is changing faster than a smoothing algorithm wants to report.

Every one of those conditions breaks one of the assumptions. That is why the technical literature keeps finding that wrist heart rate is most accurate in steady states — walking, running, cycling at a constant effort — and least accurate when intensity changes fast or the arm is doing grip work.

Apple's own engineering writeup says this out loud. Their September 2026 accuracy study explicitly chose its activity mix to include "challenging conditions for optical heart rate performance," naming cadence lock in running, loaded forearm grip in strength training and cycling, and rapid wrist motion with fast heart rate transitions in HIIT.

What the studies actually found

Two credible pieces of evidence are worth separating here, because they point in the same direction.

The independent study. A 2026 paper in Sensors put 62 healthy men through standardized endurance and resistance protocols wearing four smartwatches at once (Apple, Galaxy, Fitbit, Garmin) against ECG as the reference:

• Heart rate accuracy was high in both modalities: correlations with ECG ranged from 0.64 to 0.97, reliability was excellent (ICC above 0.94), and limits of agreement clustered around ±10 bpm.
• During endurance exercise, all four brands were comparable to ECG.
• During resistance exercise, only the Apple Watch showed no significant difference from ECG. The others drifted.
• Energy expenditure was a different story. Across all devices, calorie estimates were underestimated during endurance work and got worse during resistance training, with weak correlations (0.10-0.34) and poor reliability (ICC below 0.45).

The vendor study. Apple ran a head-to-head accuracy study published in September 2026, with 1,460 participants enrolled and 1,254 contributing paired data, across five sites, using a Polar H10 chest strap as the reference. Devices tested included Apple Watch Series 12, Garmin Forerunner 970, Google Pixel Watch 4, Huawei Watch 5, Samsung Galaxy Watch 8, and WHOOP 5, with the Oura Ring 5 on a finger. Participants did outdoor runs, indoor interval runs, cycling, HIIT and strength training.

Apple reported better accuracy than every comparison device overall and in every individual workout type, including strength training, with 139 of 159 activity-by-measure comparisons favoring Apple Watch at 99% confidence.

Two honest caveats. First, this is a study run by the company selling the product, so treat the headline the way you would treat any self-published benchmark: the methodology is unusually detailed and the reference device is a legitimate standard, but it is not neutral. Second, and more useful for your training: the Oura Ring was excluded from strength training because the manufacturer says not to wear it while lifting. That is a ring-shaped confession that wrist and finger optical sensing is genuinely hard during loaded grip work.

So the honest summary: modern optical sensing is better during lifting than the folklore says, and still not as good as it is during a run.

Where the numbers hold up and where they don't

The four reasons your mid-set number is wrong

1. Averaging lag. Your watch reports a value that reflects the recent past, not this exact instant. That is a deliberate design tradeoff: less averaging means more responsiveness but also more garbage from motion. Move from 95 bpm at the bottom of a rest period into a set that spikes you toward 160, and the watch catches up over the next several seconds — often after the set is over.

2. Grip and forearm tension. Loaded grip compresses tissue under the sensor. This is the most consistent weakness across studies, and it is why the Int J Exerc Sci comparison of a wrist device and an earbud found the wrist device's accuracy was "likely affected by wrist movements" during high-intensity intervals while the earbud held up better.

3. Watch placement and fit. A band that is loose enough to slide, or sitting too low near the wrist bone, gives a weak signal. So does wrist flexion under load. Hot take: a snug band one finger-width above the wrist bone does more for accuracy than upgrading to a newer watch.

4. The physiology is genuinely strange. Heart rate during resistance training does not behave like it does on a run. It rises steeply during a set because of sympathetic drive, circulating adrenaline and mechanical pressure, then it can stay elevated or even creep up during the rest interval rather than falling smoothly. Blood pressure rises much more during heavy lifting than during steady cardio. So part of what looks like an error is your cardiovascular system doing something cardio rules do not describe.

Which is why "my zone 1 is wrong" complaints show up so often in lifters. The watch is telling you your session sat in low zones with occasional spikes, because between sets you really are near 100-120 bpm and during sets the sensor often reports late. The average is distorted by which of those two states the algorithm happened to catch.

How to get data you can actually use during a lifting session

You do not need perfect beat-to-beat data. You need a number that is repeatable, so you can compare this week to last week.

• Start a Strength Training workout. Do not leave it on the general heart rate view. Every sampling and smoothing decision in the app is tuned for the workout you selected, and you get an HR graph afterward instead of a handful of daytime readings.
• Tighten the band before your first set and move the sensor above your wrist bone, away from the joint.
• Read your heart rate between sets, not during them. Rest 30 to 60 seconds, stand or sit still, then glance at your watch. That number is close to real, and it tells you how fast you are recovering set to set.
• Watch the recovery pattern instead of the peak. If your heart rate used to drop 25 bpm in a minute between sets and now only drops 12, that is a meaningful signal, and it is a much better one than an inaccurate peak number.
• Use a chest strap if you need precision. For anybody doing zone-based conditioning around strength work, or tracking heart rate recovery seriously, a strap is still the honest answer. It is not about brand loyalty, it is about the physics of measuring a pulse from a moving, contracting forearm.
• Accept the calorie number as a rough guess, or ignore it. This is the one place where the evidence is consistent and unflattering: energy expenditure during resistance training is the weakest measurement in the entire wearable stack. If you are using active calories to set your food intake, be aware that lifting days are probably your least accurate days.

For the deeper accuracy picture across all workout types, our breakdown of Apple Watch heart rate accuracy covers when a chest strap genuinely matters and when it does not. And if you are curious why the calorie number is the shakiest one, how accurate Apple Watch calorie tracking really is goes into the details.

What to track instead of the mid-set spike

If you want to know whether your training is working and whether you are recovering, the useful signals are mostly not from the middle of a set:

• Session-level load. Total working sets, total volume, and how hard they felt. Apple Watch training load turns this into a rolling picture you can compare week to week.
• Heart rate recovery between sets. Simple, repeatable, and sensitive to fatigue.
• RPE. A 1-10 rating of how hard a set felt. It costs nothing and it correlates with the physiology better than a lagged heart rate reading does.
• Overnight recovery. This is where the wrist sensor is genuinely excellent, because you are lying still for hours. Resting heart rate and HRV measured during sleep are the strongest signals you have. Strength training reliably suppresses HRV for 24 to 48 hours, and why strength training drops your HRV explains what the drop means and how long to expect it.
• Trend, not single nights. A one-day HRV dip after leg day is not a red flag. A two-week downward drift while your sleep is unchanged is worth acting on.

Century pulls sleep, overnight heart rate and HRV into a single daily score, which is a much more useful summary of a lifting block than trying to interpret individual mid-set spikes. You still wear the same watch. You just stop reading the noisiest number on it.

Bottom line

Your Apple Watch is better at measuring heart rate during strength training than most gym folklore claims, but it is still an optical sensor on a contracting forearm. Short, heavy sets outrun its smoothing algorithm, and loaded grip degrades the signal. Independent testing shows usable accuracy during resistance exercise, with the Apple Watch in particular tracking ECG closely. Calories during lifting remain unreliable across every brand.

So use the number that behaves: the post-set reading after 30 to 60 seconds of standing still, the trend across a block of training, and the overnight recovery data you collect while sleeping. Let the mid-set spikes be colour, not evidence. Your training decisions will get better the moment you stop treating every reading as equally trustworthy.

Century AI helps you understand your body with a daily health score, recovery score, and sleep insights — using the watch you already wear.