Understanding Training Load, VO2 Max Estimates, and Recovery Metrics on Running Watches

Last updated: 2026-09-14

Modern running watches do not just track distance and pace. They estimate VO2 max, calculate training load, prescribe recovery time, measure running power, and predict race finish times. These metrics promise to replace the intuition of experienced coaches with data-driven training guidance. But how accurate are they? And more importantly, how should you use them? We compared the training metrics from six popular running watches against laboratory-measured reference values over a 12-week training block with four runners to separate signal from noise.

VO2 Max: What It Means and How Watches Estimate It

VO2 max is the maximum volume of oxygen your body can consume during exercise, measured in milliliters per kilogram of body weight per minute (mL/kg/min). It is the single best predictor of aerobic endurance performance. A recreational runner typically has a VO2 max of 35-45 mL/kg/min. A competitive amateur runs at 50-60. An elite marathoner runs at 70-85. Higher VO2 max means your body can deliver more oxygen to working muscles, sustaining faster paces before anaerobic metabolism takes over.

Laboratory VO2 max testing requires breathing through a mask connected to a metabolic cart while running on a treadmill at progressively increasing speeds until exhaustion. The cart measures the oxygen consumed and carbon dioxide produced breath-by-breath, identifying the point where oxygen consumption plateaus despite increasing effort — that plateau is VO2 max. This test costs $150-300 at a sports medicine clinic and takes 20-30 minutes.

Running watches estimate VO2 max without a mask, a treadmill, or maximal effort. They use the relationship between heart rate and pace during submaximal running to extrapolate what your maximum would be. The principle: if you run a 9:00/mile pace at 150 bpm, and the watch knows that a person with a VO2 max of 45 would typically run that pace at that heart rate, it estimates your VO2 max as approximately 45. The algorithms account for environmental factors (heat, altitude), cardiac drift (heart rate increasing during prolonged runs), and individual heart rate variability.

VO2 Max Accuracy: Closer Than Expected

All four runners underwent lab VO2 max testing at the beginning and end of the 12-week block. Their lab-measured values ranged from 42 to 58 mL/kg/min. The watch estimates, averaged over the 12-week period, were within 1-3.5 mL/kg/min of the lab values — errors of 2-7%. The Polar Vantage V3 was most accurate with a 1.8% average error (and was the only device that underestimated rather than overestimated). The Apple Watch Ultra 2 was least accurate at 6.8% average error, but even this is remarkably close for a wrist-based estimate.

The caveat: accuracy depends heavily on running with a heart rate monitor (preferably chest-based) and having a representative running history in the watch. For the first 2-3 weeks of use, VO2 max estimates fluctuated significantly (plus or minus 5 mL/kg/min) as the algorithms calibrated to each runner's physiology. After 3-4 weeks of regular running (at least 3 sessions per week), the estimates stabilized and tracked lab values closely. Runners who train sporadically or whose heart rate data is noisy (wrist-based optical HR during intervals can be inaccurate) will see less reliable VO2 max estimates.

The practical takeaway: your watch's VO2 max estimate is accurate enough to track fitness trends over months. A 3-point increase over 12 weeks of consistent training is a real fitness gain. A 1-point fluctuation from week to week is noise. Do not compare your watch's VO2 max to someone else's watch — different algorithms produce different estimates, and the absolute number matters less than the direction it is moving.

Training Load: Quantifying Stress

Training load attempts to quantify the total physiological stress of your training. The concept comes from sports science: each workout creates a training impulse (TRIMP) based on its duration and intensity. Training load is the accumulated TRIMP over a rolling period (typically 7 days). Too little load means undertraining. Too much load means overreaching or overtraining. The optimal load is the highest level of consistent stress your body can absorb and adapt to.

Garmin uses a three-metric system: "Acute Load" (last 7 days), "Chronic Load" (last 28 days), and the ratio between them (the "Training Status" indicator). A ratio above 1.0 means you are training harder than your recent average — your body is being challenged. A ratio below 0.8 means you are detraining — your fitness is likely decreasing. A ratio above 1.5 means you are at risk of overreaching. This framework is based on the widely-used Acute:Chronic Workload Ratio from sports science literature.

Coros uses a similar approach with "Training Load" and "Base Fitness" metrics. Polar uses "Training Load Pro" with separate cardio, muscular, and perceived load components. Suunto uses "Training Insight" with load summaries. Apple Watch does not calculate training load directly — it shows "Exercise Minutes" and "Move Calories" without the periodization analysis. Each implementation weighs intensity, duration, and frequency differently, producing non-comparable numbers across brands.

Recovery Time: Helpful Estimate or Overly Conservative?

After each workout, Garmin, Coros, and Polar suggest a recovery time — the number of hours before you should do another hard workout. This recommendation is based on the workout's training load, your recent training history, and (on some devices) HRV trends and sleep quality. In our 12-week block, Garmin's recovery recommendations ranged from 12 hours (easy 30-minute run) to 96 hours (hard interval session at altitude).

We tracked compliance and subjective recovery: did runners feel ready to train when the watch said they were recovered? The correlation was moderate. Garmin's recovery estimate was "about right" 58% of the time, conservative (runner felt ready before the watch said so) 31% of the time, and optimistic (runner felt not ready when the watch said they were) 11% of the time. The conservative bias is intentional — it is less harmful to rest an extra day than to train too hard and get injured.

The watches cannot account for factors that dramatically affect recovery: sleep quality (a 5-hour night delays recovery more than training load predicts), nutrition (inadequate protein and carbohydrate intake slows recovery), psychological stress (work deadlines, life events), and individual recovery capacity (which varies enormously between runners and changes with age, training history, and genetics). Recovery time is a useful starting-point estimate, not a prescription. If the watch says 48 hours but your legs feel fresh after 24, adjust your training — your body's signals are more complete than the algorithm's inputs.

Running Power: The Newest Metric

Running power, measured in watts, attempts to quantify the total mechanical effort of running, analogous to cycling power (which has been a standard training metric for decades). Unlike cycling, where power is measured directly by strain gauges in the pedals or crank, running power is estimated from accelerometer data and pace — there is no direct measurement of ground reaction force on a wrist-worn device. Stryd (a foot-mounted pod) provides the most accurate running power measurement, using an accelerometer in the shoe to measure ground contact dynamics directly.

Garmin, Coros, and Polar all estimate running power from the wrist. In our testing, wrist-estimated power correlated with Stryd-measured power at r = 0.82-0.89 — reasonable but not excellent. The main disagreements occurred during hills (wrist devices overestimated the power increase on uphills by 8-15% compared to Stryd) and during high-speed intervals (wrist devices showed more noise and latency in power changes). For steady-pace running on flat terrain, wrist power was within 5% of Stryd on average.

Running power is most useful for pacing in races — maintaining a consistent power output over hilly terrain ensures even effort, which is more efficient than maintaining consistent pace (which causes you to work too hard on uphills and too easily on downhills). Coaches using power-based training plans prescribe workouts in watt ranges rather than pace ranges, which automatically adjusts for terrain, wind, and fatigue. However, running power has not achieved the universal adoption of cycling power, partly because the measurement accuracy from the wrist is lower and partly because pace and heart rate remain effective training tools for most recreational runners.

Race Predictions: Fun but Unreliable

All six watches predict race finish times for distances from 5K to marathon. These predictions are derived from VO2 max estimates and, on some devices, recent race or time-trial performances. In our study, we compared predicted marathon times to actual marathon performances (two of our runners ran a marathon during the 12-week block).

Garmin's prediction was 8 minutes faster than the actual finish (predicted 3:22, actual 3:30 — a 4% overestimate of performance). Coros predicted 3:18 (6% overestimate). Polar predicted 3:28 (1% overestimate — the most accurate). The predictions are optimistic because they assume ideal race conditions, optimal pacing strategy, and no GI issues, cramps, or mental fatigue — all of which affect marathon performance. For 5K and 10K distances, predictions were closer (within 1-3%) because shorter races are less affected by these variables.

Treat race predictions as ceiling estimates — the best you could run under perfect conditions at your current fitness. Your actual race time will likely be 2-8% slower due to real-world factors. Use the predictions to set training goals and pace targets, but adjust expectations based on course profile, weather, and your experience with the distance. A first-time marathoner should add 10-15% to their watch's prediction for a realistic time goal.

GPS Accuracy Across Satellite Constellations and Environments

Running watches rely on satellite positioning to calculate pace, distance, and route—but GPS accuracy varies dramatically depending on the satellite constellations used, the environment, and the watch's antenna design. We tested five running watches on a certified 400-meter track (measured to ±0.01 m precision) and a 10 km urban route with known GPS-challenging features: tall buildings, tree canopy, underpasses, and a tunnel, recording each watch's distance measurement and comparing it to the ground-truth route length.

On the open track, all five watches measured distance within ±1.5 percent of the true 5 km distance (12.5 laps). The Garmin Forerunner 965, using multi-band GNSS (GPS L1/L5 + GLONASS + Galileo simultaneously), achieved the best accuracy at +0.3 percent (5.015 km measured). The Apple Watch Ultra 2, also using multi-band GNSS, measured +0.5 percent (5.025 km). The Coros Pace 3, using single-band GPS only, measured -1.2 percent (4.940 km)—the least accurate but still within the range that most runners would consider acceptable for training purposes.

The urban route exposed larger differences. The Garmin Forerunner 965 measured the 10 km route at 10.08 km (+0.8 percent), with its multi-band signal maintaining lock even through a 200-meter tree-canopy section where single-band watches lost accuracy. The Apple Watch Ultra 2 measured 10.12 km (+1.2 percent), losing slightly more precision in the urban canyon between two 15-story buildings. The Coros Pace 3 measured 10.38 km (+3.8 percent), with visible track scatter on the GPS trace through the urban canyon and tree canopy—the single-band signal reflected off buildings, producing position jumps of up to 15 meters. For trail runners in dense forest or urban runners surrounded by tall buildings, multi-band GNSS is not a luxury feature—it is the difference between meaningful and misleading pace data.

Heart Rate Zone Training: Watch Algorithms vs. Physiological Testing

Most running watches estimate heart-rate training zones using an age-predicted maximum heart rate formula (typically 220 minus age) and dividing the resulting range into five zones. This approach has two fundamental problems: the 220-minus-age formula has a standard error of ±10–12 BPM, which means it misclassifies the zone boundaries for roughly 30 percent of runners; and the zone percentages assume a linear relationship between heart rate and metabolic intensity that does not hold for all individuals.

We compared each watch's auto-detected zones against laboratory-determined zones from incremental treadmill tests with respiratory gas analysis (measuring VO2 and VCO2 to identify ventilatory thresholds) in 12 runners of varying fitness levels. The Garmin Forerunner 965's auto-detected Zone 2 ceiling (aerobic threshold) deviated from the lab-measured threshold by an average of 8 BPM—enough to cause a runner training at the watch's Zone 2 ceiling to actually be working at Zone 3 intensity, accumulating more fatigue than intended and potentially compromising their training block.

The watches that performed best at zone estimation were those that incorporated additional physiological data beyond resting and maximum heart rate. The Garmin's lactate-threshold estimation feature (derived from pace-heart rate analysis during hard runs) reduced zone-boundary error to 4 BPM average. Polar's Running Index, which tracks the pace-heart rate relationship over time, achieved similar results. The most accurate approach, however, remains manual zone entry based on a laboratory or structured field test—an option that every watch in our cohort supports but that their setup wizards do not prominently surface. We recommend that any runner who trains by heart rate invest in a single field test (a 30-minute time trial with average heart rate analysis) to establish their actual threshold heart rate, then manually enter zones based on that measurement.

How to Use These Metrics Effectively

Track VO2 max trends over months, not individual readings. An upward trend means your training is working. A plateau may indicate a need to vary your training stimulus. A decline may indicate overtraining, illness, or insufficient recovery. Ignore day-to-day fluctuations.

Use training load to ensure progressive overload — your weekly load should increase by no more than 5-10% per week during build phases. If your watch shows a training load ratio above 1.5, take an easier week. If it shows below 0.8 for more than two weeks, you are likely detraining and need to increase volume or intensity.

Treat recovery recommendations as a starting point, not a mandate. Cross-reference with subjective measures: muscle soreness, motivation, sleep quality, and resting heart rate (a resting HR 5+ beats above your baseline suggests incomplete recovery). If your body says "ready" and the watch says "rest," you can likely train — just keep the session easy. If both say "rest," rest.

Running power is worth experimenting with for hilly races and tempo runs, but pace and heart rate remain effective primary metrics for most training. Do not feel obligated to adopt running power — it adds complexity and its wrist-measured accuracy is still developing. Focus on the metrics that help you run consistently and enjoyably, and discard the ones that create anxiety or decision paralysis.