How Accurate Is Wrist Optical Heart Rate During Lifting?

Your watch says 182 on the third rep of a heavy set, but you're breathing fine and the bar speed hasn't dropped off. Rack it, glance down, and the number is still climbing — past what your chest strap read during an actual sprint test months ago. Your heart isn't malfunctioning. The sensor reading it is.

Wrist optical heart rate — the green-LED photoplethysmography (PPG) built into nearly every fitness watch since roughly 2015 — was validated mostly on steady, rhythmic activity: walking, running, cycling. Grip a barbell hard enough to whiten your knuckles and the physics the sensor depends on changes underneath it. The question that matters isn't whether wrist optical HR is accurate in general — it's where specifically it breaks down under load, and what to trust instead when it does.

How Wrist Optical Heart Rate Actually Works

A PPG sensor shines green light (infrared, too, on newer devices) into the skin and measures how much bounces back to a photodiode. Blood absorbs more light than surrounding tissue, so every heartbeat's pulse through the capillary bed under the sensor makes the reflected light dip slightly. That rhythmic dip-and-recover pattern, sampled dozens of times per second, is the raw signal the watch converts into a beats-per-minute number.

The catch: the raw signal is noisy, mostly from motion — your arm swinging, your wrist flexing, the case shifting against skin. Manufacturers correct for this with motion compensation: an accelerometer records your wrist's movement, and an algorithm subtracts frequencies matching your stride from the optical signal, leaving just the cardiac frequency. That works when motion is periodic, which a running stride is. It works far worse when the wrist isn't swinging in any clean rhythm — exactly the situation mid-set.

Why Grip and Motion Break the Signal During Lifting

Two failure modes stack during resistance training, and either alone is enough to throw off a reading.

The first is mechanical. Gripping a bar hard compresses the vessels and tendons running directly under the sensor, shrinking the optical signal's amplitude. Some algorithms respond to a weak signal by holding the last confident reading, which is why a watch will sometimes freeze on one number for 10-15 seconds through a heavy set, then jump once grip pressure releases.

The second is a motion problem specific to lifting's rep structure. A barbell rep isn't smooth and periodic the way a stride is — it has a concentric phase, a pause, and an eccentric phase, often at 40-70 reps per minute, which overlaps resting and light-exercise heart rate frequencies. When the accelerometer-derived motion frequency lands close to the true pulse frequency, the algorithm can lock onto rep cadence instead of heartbeat — a failure mode sometimes called cadence lock. In that moment the watch isn't reporting your heart rate at all; it's reporting bar speed and mislabeling it as BPM.

What the Accuracy Studies Actually Found

Gillinov et al. (2017), in Medicine & Science in Sports & Exercise, fitted 50 healthy adults with an ECG reference alongside six wrist monitors across treadmill walking, treadmill running, outdoor cycling, and elliptical training. The Apple Watch had the lowest overall error, but even it climbed from roughly 2% mean absolute percentage error on the treadmill to nearly 7% on the elliptical; the other five devices fared worse, some past 20%. What mattered wasn't jostling — it was whether the activity involved continuous hand-gripping. The study didn't test resistance training directly and excluded anyone with a diagnosed arrhythmia, so its numbers describe grip-adjacent cardio, not a barbell.

Boudreaux et al. (2018), also in Medicine & Science in Sports & Exercise, closed that gap by testing four wrist monitors during real resistance exercise — squat, bench, and deadlift — against a chest-strap ECG reference in trained lifters. Error was substantially higher than the same devices produced cycling in the same session; several exceeded 20% mean absolute error on the free-weight sets, and none cleared the CTA-2065 industry benchmark of 10% during resistance exercise, despite clearing it on the bike. The tradeoff: a single session at a fixed, moderate load — it doesn't say whether error worsens as grip fatigue builds late in a workout.

Which Lifts Fail Worst — and Which Are Fine

The pattern across both published data and informal chest-strap testing in a normal gym is consistent: error tracks grip intensity and hold duration, not muscle group or load.

• Worst: Deadlifts (especially the lockout hold), farmer's carries, static holds, heavy pull-ups or rows — a sustained crushing grip directly over the vessels the sensor reads.
• Moderate: Bench and overhead press — grip is firm but not maximal, and the cyclical arm path at least gives the accelerometer a learnable pattern.
• Closer to reliable: Leg press, leg extension, hip thrust — hands rest lightly rather than gripping under tension.

A program using heart-rate autoregulation for machine days but heavy pulling upper-body needs two monitoring strategies, not one blanket rule about the watch.

A Field Protocol to Test Your Own Watch Against a Chest Strap

You don't need a lab to find out whether your watch holds up during your lifts. Pair a chest strap (Polar H10, Garmin HRM-Pro, or similar) to a phone app that logs continuous BPM, and run your watch on the wrist you normally train with.

1. Baseline, 5 min seated: Both devices should agree within 2-3 bpm, or recheck watch fit first.
2. Warm-up, 5 min light cardio: Readings should track within roughly 5% — the modality where wrist optical HR sits closest to rated accuracy.
3. Working sets, 3 x 5 at 75-80% 1RM: Note the watch's and strap's BPM at the top of the last rep, timestamped within the same 2-3 second window.
4. Recovery at 15s, 30s, 60s post-rack: Compare the decay curve, not single points — a lagging wrist monitor declines slower and smoother than the strap's true curve.

Calculate percent difference as |wrist minus chest| divided by chest, times 100. Below 5% is close enough to trust for trends. Between 5% and 12%, treat it as directional only — useful for noticing recovery feels slower, not for a hard rest-period cutoff. Above 12-15%, especially on grip-dominant lifts, the wrist reading is measuring rep cadence, not your heart.

When to Switch to a Chest Strap

Wrist optical HR is genuinely fine for weekly trends on recovery days, steady-state cardio, and a rough sense of whether today's resting rate runs higher than usual. None of that needs second-by-second precision.

It stops being fine the moment a watch number gates a real-time decision: cutting a set short because the display spiked, extending rest because a number hasn't come down when it's just frozen, or setting an EMOM's next round off a BPM that's really measuring grip. Any of those calls for a chest strap or ECG-based armband, which read electrical activity rather than blood-volume changes and are far less sensitive to grip — though even a strap can lose contact if sweat pools under the electrodes, so an implausible number is still worth checking against a manual pulse count.

References

1. Gillinov, S., Etiwy, M., Wang, R., et al. (2017). Variable Accuracy of Wearable Heart Rate Monitors during Aerobic Exercise. Medicine & Science in Sports & Exercise, 49(8), 1697-1703.
2. Boudreaux, B.D., Hebert, E.P., Hollander, D.B., et al. (2018). Validity of Wearable Activity Monitors during Cycling and Resistance Exercise. Medicine & Science in Sports & Exercise, 50(3), 624-633.
3. Spierer, D.K., Rosen, Z., Litman, L.L., & Fujii, K. (2015). Validation of Photoplethysmography as a Method to Detect Heart Rate During Rest and Exercise. Journal of Medical Engineering & Technology, 39(5), 264-271.