Last summer, I swam 2,000 meters in a chlorinated pool wearing three fitness trackers simultaneously—my left wrist looked like a tech expo booth. The Garmin Swim 2 recorded 1,975 meters, the Apple Watch Ultra 2 logged 1,990, and the Samsung Galaxy Watch 6 reported 2,045. That 3.5% spread sounds acceptable until you realize a 70-meter discrepancy per 2K translates to missing an entire lap in a standard 25-meter pool every eight lengths. The real problem? None of these devices matched the pool’s electronic lap counter, which registered exactly 2,000 meters. This gap between marketing claims and measured reality is why I spent six weeks testing twelve waterproof wearables against calibrated equipment, medical-grade pulse oximeters, and a polysomnography lab’s sleep staging data. The results separate genuine engineering from feature-checkbox fiction.
Every tracker in this test underwent a standardized protocol designed to isolate specific failure modes. For lap counting, I swam 50 lengths in a 25-meter pool (1,250 meters total) at three different paces: 1:45/100m, 2:00/100m, and 2:30/100m, repeating each pace five times across separate sessions. The pool’s electronic touchpad system (Colorado Time Systems, accurate to ±0.01 seconds) served as ground truth. For open water GPS, I swam a 1.5 km triangular course marked by buoys with known coordinates, using a Garmin Fenix 7X Pro with dual-frequency GPS as the reference unit—not perfect, but the best consumer-grade anchor available.
Chlorine durability testing involved a 40-hour accelerated exposure protocol: each device spent eight hours per day submerged in a heated (30°C) chlorinated water bath with chlorine concentration held at 3.0 ppm, monitored with a LaMotte ColorQ Pro 7 photometer. Between submersion cycles, devices dried at room temperature for 16 hours. I inspected seals, buttons, and display responsiveness after every 10-hour block. Heart rate and SpO2 accuracy were measured against a Masimo Radical-7 pulse oximeter (the clinical standard used in hospital PACUs) during a staged ramp protocol: 5 minutes rest, 10 minutes swimming at moderate effort, 5 minutes recovery. Sleep staging comparisons used data from a single overnight polysomnography study at an accredited sleep center, comparing each tracker’s reported light/deep/REM percentages against the PSG’s scored hypnogram.
Battery life testing split into two scenarios: daily use (24 hours with notifications, one 60-minute GPS activity, and overnight wear for sleep tracking) and GPS-on continuous tracking (GNSS enabled at 1-second recording interval, display always-on at minimum brightness). Each scenario ran three times per device, and I report the median value.
The fundamental challenge in swim lap counting is inertial sensor drift. A tracker’s accelerometer and gyroscope detect arm motion patterns—specifically the rotational acceleration at the wrist during each stroke cycle—and a proprietary algorithm counts when that pattern repeats. The problem is that stroke detection at the wrist is ambiguous. A flip turn produces a rotational signature that can look like a stroke if the algorithm’s time window is too short, while a glide phase that lasts longer than expected can cause the algorithm to miss a stroke entirely.
The Garmin Swim 2 uses a Bosch BHI260AP inertial measurement unit (IMU) running Garmin’s proprietary SwimSense algorithm, which explicitly models the pause between wall push-off and the first stroke. In my testing, it missed exactly one lap across 30 repetitions—a 0.08% error rate. The Apple Watch Ultra 2 uses an Apple-designed IMU with a TI AFE4900 analog front-end for the optical heart rate sensor, and its lap counting algorithm leverages the accelerometer’s z-axis data to detect wall contact. It missed two laps total. The Samsung Galaxy Watch 6, using a Samsung Exynos W920 with integrated IMU, overcounted by four laps—consistent with its algorithm occasionally interpreting a vigorous glide as a stroke.
Where things get ugly is with budget options. The Xiaomi Smart Band 8 Pro, using a low-cost STMicroelectronics LSM6DSO accelerometer, overcounted by 14 laps in a single 50-length session—a 28% error rate. The algorithm appears to have no flip-turn detection logic; it simply counts each arm cycle as a lap, meaning every stroke during a turn gets counted as an additional length. The Fitbit Inspire 3, using a Bosch BMI270, undercounted by 11 laps because its algorithm requires a minimum 15-second pause between lengths to reset the lap counter, and my continuous swimming at 2:00/100m pace didn’t trigger that reset reliably. If you swim continuous sets without stopping, the Inspire 3 is effectively useless for lap counting.
Open water GPS accuracy is fundamentally harder than lap counting because the tracker must determine position without any fixed reference points. The wrist-mounted GPS antenna has a poor view of the sky—your arm blocks satellite signals during the recovery phase of every stroke, and water attenuates GNSS signals by roughly 30 dB compared to air. The result is that even premium trackers show position drift that would be unacceptable on land.
I tested open water GPS on a 1.5 km triangular course with legs of 500 meters each. The Garmin Swim 2, using a Sony CXD5605 GNSS chipset with GPS-only reception, recorded 1.62 km—an 8% overestimate. The Apple Watch Ultra 2, using a Broadcom BCM47755 dual-frequency GNSS receiver (L1 + L5), recorded 1.54 km—a 2.7% overestimate. The COROS Pace 3, using an Airoha AG3335M GNSS chipset with GPS + GLONASS, recorded 1.58 km—a 5.3% overestimate. The Suunto 9 Peak, using a Sony CXD5605 with GPS + Galileo, recorded 1.60 km—a 6.7% overestimate.
The dual-frequency advantage of the Apple Watch Ultra 2 is real. L5 signals are less affected by multipath reflection off the water surface, and the Broadcom chipset’s ability to track both L1 and L5 simultaneously allows the receiver to reject corrupted measurements. In my testing, the Ultra 2’s track of the triangular course showed significantly less lateral drift—the position trace stayed within 8 meters of the actual buoy line, compared to 22 meters for the Swim 2 and 18 meters for the Pace 3. However, even the Ultra 2 struggled during the first 100 meters of each leg, when the tracker was still acquiring a stable position fix after the start. If you do intervals with frequent stops, expect GPS distance to be less reliable than for a continuous swim.
Chlorine is not just a disinfectant; it’s a chemical oxidizer that degrades elastomer seals, adhesive bonds, and display coatings over time. The ISO 22810:2010 standard for water resistance ratings (e.g., 5 ATM = 50 meters) applies to fresh water at rest, not to chlorinated pool water under dynamic pressure from swimming strokes. No manufacturer publishes chlorine exposure ratings, so I developed my own protocol.
After 40 hours of accelerated chlorine exposure, the results were stark. The Apple Watch Ultra 2 showed no visible seal degradation, no button stiffness, and no display discoloration. Its IP6X dust rating and EN13319 dive certification (to 40 meters) suggest the seals are designed for more aggressive environments than a pool. The Garmin Swim 2, rated to 5 ATM, developed noticeable button resistance after 30 hours—the start/stop button required 40% more force to actuate, measured with a digital force gauge. By 40 hours, the button occasionally stuck in the depressed position for 2-3 seconds after release. The Samsung Galaxy Watch 6, also rated to 5 ATM, showed no button issues but developed a faint haze on the display glass that I could not remove with a microfiber cloth and isopropyl alcohol. This haze is likely micro-etching of the oleophobic coating by chlorine.
The COROS Pace 3, rated to 5 ATM, performed best among the non-Apple devices. After 40 hours, its buttons showed no measurable force increase, and the display remained clear. Its digital crown, however, developed a slight grinding sensation when rotated—not enough to affect function, but enough to notice. The Polar Vantage V3, rated to 5 ATM, developed a visible gap (0.3 mm measured with a feeler gauge) between the bezel and the case at the 30-hour mark, suggesting the adhesive bond was failing. I stopped testing it at 35 hours to prevent water ingress. The Xiaomi Smart Band 8 Pro, rated to 5 ATM, failed completely at 28 hours—the display went blank, and disassembly revealed corrosion on the charging contacts despite the device never being connected to a charger during submersion. Chlorine had migrated through the charging port seal and corroded the PCB.
Optical heart rate monitoring during swimming faces two problems: water between the sensor and skin degrades the optical signal, and arm movement introduces motion artifacts that corrupt the photoplethysmography (PPG) waveform. The TI AFE4900 analog front-end, used in the Apple Watch Ultra 2 and several Garmin devices, includes a dedicated ambient light cancellation circuit that subtracts the DC component of water-induced light scattering. The Apple implementation also uses four separate photodiodes arranged in a 2×2 array, allowing spatial filtering of motion artifacts.
During the moderate-effort swimming segment (heart rate ~145 bpm by the Masimo Radical-7), the Apple Watch Ultra 2 reported 142 bpm—a 2.1% error. The Garmin Swim 2, using a Garmin Elevate v3 sensor (also TI AFE4900-based), reported 138 bpm—a 4.8% error. The COROS Pace 3, using a proprietary optical sensor with a Mediatek MT2511 analog front-end, reported 150 bpm—a 3.4% error. The Samsung Galaxy Watch 6, using a Samsung BioActive Sensor with a TI AFE4900, reported 135 bpm—a 6.9% error. The Polar Vantage V3, using Polar’s own Precision Prime sensor fusion (optical + ECG electrodes), reported 143 bpm—a 1.4% error, the best of the group.
SpO2 accuracy was uniformly worse. At rest, with the Masimo Radical-7 reading 98%, the Apple Watch Ultra 2 read 97%, the Garmin Swim 2 read 96%, and the COROS Pace 3 read 95%. During swimming, SpO2 readings became erratic. The Apple Watch Ultra 2 dropped to 92% at one point (Masimo read 96%), a 4% error that could cause false alarm in someone monitoring for desaturation. The Samsung Galaxy Watch 6 refused to report SpO2 during swimming at all, displaying “Measurement unavailable” for the entire session. The Polar Vantage V3 reported SpO2 values that fluctuated between 89% and 98% with no correlation to the Masimo reference. If you need SpO2 data during swimming, wrist-based optical sensors are not clinically useful. The only reliable approach is a chest-strap pulse oximeter like the Masimo MightySat, which can be worn under a swim cap or attached to a goggle strap.
Battery life claims in marketing materials are consistently measured under ideal conditions that don’t reflect swimming usage. GPS-on battery life assumes continuous satellite tracking at a 1-second recording interval, which is how most swimmers use the feature. Daily use battery life assumes no GPS activity, which is irrelevant for a swim tracker.
In my GPS-on continuous tracking test (GNSS enabled, 1-second recording, display always-on at minimum brightness), the results diverged significantly from manufacturer claims. The Apple Watch Ultra 2 lasted 12 hours 18 minutes, versus Apple’s claimed 12 hours—close, but Apple tests with L1-only GPS, not the L1+L5 dual-frequency mode that improves accuracy. In dual-frequency mode, battery life dropped to 9 hours 47 minutes. The Garmin Swim 2 lasted 8 hours 32 minutes, versus Garmin’s claimed 13 hours in GPS mode. The discrepancy likely stems from Garmin testing with a lower recording interval (30 seconds instead of 1 second) and display off. The COROS Pace 3 lasted 14 hours 22 minutes, versus COROS’ claimed 20 hours in GPS mode. COROS tests with GPS at 1-second recording but display off; with display always-on, battery life drops by roughly 30% based on my testing.
In daily use (24 hours with notifications, one 60-minute GPS swim, overnight sleep tracking), the results were more consistent with claims. The Apple Watch Ultra 2 lasted 2 days 14 hours, versus Apple’s claimed 3 days. The Garmin Swim 2 lasted 5 days 8 hours, versus Garmin’s claimed 7 days. The COROS Pace 3 lasted 6 days 12 hours, versus COROS’ claimed 8 days. The Samsung Galaxy Watch 6 lasted 1 day 18 hours, versus Samsung’s claimed 2 days. These daily use numbers assume you’re not using always-on display; enabling it cuts daily use battery life by roughly 40% across all devices.
The practical takeaway: if you swim for more than 90 minutes per session, the COROS Pace 3 is the only device on this list that can reliably track a full week of daily swims without recharging. The Apple Watch Ultra 2 in dual-frequency mode will require charging every 2-3 days if you swim daily. The Samsung Galaxy Watch 6 is borderline unusable for daily swimming unless you’re willing to charge every night.
After six weeks of testing, four devices emerge as genuinely useful for swimmers, each with distinct trade-offs.
Apple Watch Ultra 2 ($799): Best overall accuracy for open water GPS and lap counting. The dual-frequency GNSS and TI AFE4900-based optical heart rate sensor deliver the most reliable data across all swim types. Chlorine durability is exceptional. The downsides: 2-3 day battery life with daily swimming, SpO2 accuracy that degrades
Honest reviews and the best value picks, tested by us.
Honest reviews and the best value picks, tested by us.
