Swim watches count steady front-crawl pool lengths well, but stroke counts, wrist heart rate in water and open-water distance are unreliable or unvalidated
Interpretation
Consistent direction across several small device-validation studies of different years and device types. Pool distance for steady front crawl was mostly correct (7 of 9 watches always correct over 400 m in 2023 testing, Jacko 2024; no significant distance error in 36 adults, Budig 2021, and 10 national-level swimmers, Mooney 2017), but stroke changes (errors up to 91.7%), irregular technique (+500 m on a 500 m swim) and first/last laps cause errors; stroke counts were out by 0.4-29.5% (Jacko 2024) and differed from video in most tests (Mooney 2017); a head-worn sensor underestimated distance (Shell 2021). Wrist optical heart rate in water read low with LoA -53.0 to 33.9 bpm in 26 swimmers (Olstad 2020) and deviation rose 13-56 bpm in 10 swimmers (Cosoli 2022); temple sensors agreed better (Olstad 2020; Kits 2026, MAPE 1.0%). Open-water evidence is limited (Seelhöfer 2026) and no wrist-watch open-water distance validation was found. Gaps: small samples, mostly competitive swimmers, devices date quickly.
Practical implication
Trust pool lengths for steady front crawl, mark drills and pauses manually, use stroke metrics only as same-device trends, set swim intensity by pace rather than wrist heart rate, use a temple sensor or a well-secured chest strap for heart rate, and treat open-water distance as an estimate. Check your own device against counted lengths, a chest strap and a known course, because findings date as firmware changes.
Sources
9 sources, independently verified.
400 m front crawl: 7 of 9 watches recorded correct distance in all three trials; 2 had MAPE 4.2% (one 450 m reading each). 200 m medley with stroke changes every 50 m: 6 correct, others 8.3%, 16.7% and 141.7%. Medley with changes every 25 m: none correct, MAPE 41.7-91.7%. Strokes per 50 m (5 watches): MAPE 0.4-29.5%. Running heart rate MAPE 3.1-8.3% (not tested in water). Devices of 2023-24; firmware changes date the findings.
500 m and 1,000 m pool swims at 2:00-2:15/100 m: swim distance not significantly different from criterion overall, but a maximum deviation of +500 m on the 500 m swim for one participant with poor technique. Wrist heart rate over 500 m: MAPE 3.29% and 8.61% for the two watches. Over 1,000 m the chest-strap reference slipped and was not valid (apparent MAPE up to 55.32%). Two 2021 multisport watches. Full text read on PMC.
Two wrist-worn swim monitors vs video over a 1,500 m four-stroke protocol: stroke type identified, no significant difference for swim distance, mid-interval lap times accurate but first and last laps less so; stroke counts differed significantly in all but two occasions, affecting stroke rate, stroke length and speed. Authors expect larger errors in recreational swimmers. Older devices.
Head-worn 9-axis inertial sensor vs video: underestimated total distance; larger absolute errors for stroke count and velocity; misclassified time in each stroke by 51% (SD 15%) and did not detect drill. Good test-retest reliability but not valid for distance, stroke count, velocity or stroke type. One 2021 device.
GNSS sampling 1-15 Hz, heterogeneous devices and reporting quality; strength of evidence limited. Proposes a framework for continuous GNSS-based triathlon swim analysis because such data are lacking. No validation of wrist-watch open-water swim distance.
Integrated GPS-accelerometer unit vs video over 100 m butterfly, breaststroke and freestyle: strong stroke-count agreement for breaststroke and butterfly; GPS mid-pool velocity not different from video for freestyle and breaststroke (95% LoA 0.36 m/s freestyle); relative errors 10.2-13.4% across strokes. Not a wrist watch; pool, not open water.
During a training session with varied intensities, against a chest strap under the swimsuit: the wrist optical watch gave significantly lower HRmax and HRmean, poor-to-good ICCs, limits of agreement -53.0 to 33.9 bpm; a temple-mounted optical sensor under the cap showed no difference and LoA -26.9 to 24.7 bpm. Notes chest straps slide down, especially on men without a swimsuit over the chest. Full text read on PMC. 2020-era devices.
Two wrist smartwatches vs chest strap on land and in water: precision and accuracy worsened in swimming, with absolute increases in deviation of 13-56 bpm (mean) and 49-52 bpm (SD); performance worsened with intensity. Water and arm movement act as interference. 2022-era devices.
Temple-mounted optical sensor in goggles vs a swim chest strap during an assigned workout: mean HR 136 ± 17 vs 135 ± 18 bpm, MAPE 1.0 ± 1.2%; concordance 0.91 (intervals) and 0.99 (whole workouts); fixed bias about -0.2 bpm. One device type, one study.