What your watch knows and what it guesses

The watch on my wrist says I need 48 hours of recovery. It says my training load is optimal. It estimates my VO2 max at 52.

The recovery time is the most invented of the three. Training load is at least derived from heart rate and duration, which are real measurements, even if the formula applied to them is debatable. The VO2 max estimate is inferred from pace and heart rate, which is a real signal processed through a rough model. The recovery countdown is a guess built on top of guesses.

This matters for anyone who trains more than one way. A watch optimized for running will give a reasonable recovery estimate after a 10K. Add a heavy squat session that afternoon, and the watch either ignores it entirely or guesses. The recovery recommendation doesn't change. The training load doesn't reflect what actually happened. The number on the screen is fiction that looks like data.

The part that works

Heart rate data from a chest strap is a solid record of how hard the cardiovascular system worked. Decades of research back this up. Wrist-based optical sensors are reasonable for steady-state running but notably less accurate during strength training and high-intensity intervals, where wrist motion and grip pressure introduce noise.

Duration is real. Distance from GPS is real in open terrain, within a few percent. On tight switchbacks or under heavy canopy, it can be off by more.

Elevation data from a barometric altimeter is real. From GPS alone, it's rough but usable.

An 8-kilometer run in 45 minutes with an average heart rate of 155. That's measured. It's a record of what happened.

The part that's guessed

Training load is where things get creative.

Most watches compute a single training load number using a method based on heart rate and time. The specific formula varies by manufacturer, but the idea is the same. Longer and harder sessions produce higher numbers. This is reasonable for endurance activities. It's what the research supports.

The problem arrives with anything that doesn't raise heart rate in the same way. A heavy deadlift session might keep average heart rate around 120. The watch sees 45 minutes at a low heart rate and scores it as an easy session. The posterior chain disagrees.

Some watches now include strength tracking. They detect wrist motion, count something resembling reps, and add a number to the training load. This is better than ignoring the session entirely. But the number is not based on the same validated methods as the endurance load. It's the manufacturer's best guess at what the session cost the body, derived from motion data and heart rate.

The issue is that these numbers are presented the same way. Same scale, same color coding, same dashboard. There's no indication that the running load is backed by 50 years of research and the strength load is an estimate based on wrist accelerometer data. They look equally trustworthy. They aren't.

Recovery estimates

Recovery time predictions are the most confident-looking number on the screen and the least reliable.

The calculation typically uses recent training load, estimated fitness level, and the difficulty of the session just finished. Some models factor in sleep and heart rate variability. The output is a countdown. 36 hours. 52 hours.

For a runner who only runs, this can be directionally useful. It captures the general idea that a hard long run needs more recovery than an easy 5K. The specific number is still a guess, but the ordering is right.

For someone who runs and lifts, the number breaks down. The watch doesn't know that yesterday's squats destroyed the legs now being used for an easy run. It doesn't know that the "rest day" recovery estimate should be different because of 160 kilos of deadlifts that morning. It models one kind of fatigue and ignores the other.

VO2 max estimates

The watch's VO2 max number is derived from the relationship between pace and heart rate during running. Faster at a lower heart rate pushes the estimate up. Slower at a higher heart rate pushes it down.

This is a reasonable proxy for aerobic fitness over time for runners. It tracks trends. An estimated VO2 max climbing from 48 to 52 over six months probably means running fitness has improved.

The number itself should not be taken literally. Lab-tested VO2 max and watch-estimated VO2 max can differ by 5 to 10 percent or more. The watch has no way to measure oxygen consumption. It's inferring from pace and heart rate, which are related to VO2 max but are not the same thing.

For someone who does both endurance and strength work, the number can be noisy in ways that are hard to interpret. A heavy lifting week might slightly elevate resting heart rate, which could make the watch think aerobic fitness dropped. It didn't. The body is just fatigued from squats.

What to actually trust

The useful rule: trust recorded data, hold estimated data loosely.

Recorded: heart rate, duration, distance, elevation, GPS track, pace. These are measurements. They happened.

Estimated: training load, recovery time, VO2 max, training status, energy scores. These are computed from models. The models are better for some activities than others. They're almost always built for endurance first and retrofitted for everything else.

For one sport, the estimates are probably directionally right. For several, the estimates are blending things that don't blend well. The raw data is still useful. The heart rate from a squat session is real. The duration is real. What the watch does with that data afterward is where it starts guessing.

The honest approach is to use the data the watch records and be deliberate about what it means. Duostride computes two separate loads, one for endurance and one for strength, each using the methods that are actually validated for that kind of work. Not because it's simpler. Because it's the only version that isn't making things up.