HRV for Athletes: What It Measures and What It Cannot

What it is measuring

The variation in time between consecutive heartbeats. A healthy autonomic nervous system produces beats that are not perfectly regular — the interval fluctuates with breathing and with parasympathetic tone.

Higher variability generally reflects greater parasympathetic ("rest and digest") influence. Lower variability reflects sympathetic dominance: stress, fatigue, illness onset, alcohol, poor sleep, or the normal aftermath of a hard session.

It is a proxy for autonomic state, and autonomic state is a proxy for recovery. Two layers of proxy, which is why the trend matters and the individual reading does not.

SDNN and RMSSD

Two statistics computed on the same beat-to-beat intervals.

RMSSD — root mean square of successive differences — is more sensitive to short-term parasympathetic activity and is what most training-focused platforms use.

SDNN — standard deviation of normal-to-normal intervals — captures more total variability across a window.

Apple Watch reports SDNN, measured overnight, which is why Base uses SDNN: consistency with the source matters more than the choice between the two statistics. Converting or re-deriving would introduce error to gain a theoretical preference. What matters far more than which statistic you use is that you use the same one, measured the same way, at the same time of day, every day.

Why the baseline is everything

An HRV of 48 ms tells you nothing on its own. Against a thirty-day mean of 45 with a standard deviation of 6, it tells you today is mildly above normal.

Base expresses it as a z-score against your own thirty days and weights it at 50% of recovery score:

zHRV   = (todayHRV - mean30d) / sd30d
zRHR   = (mean30d - todayRHR) / sd30d      // inverted: lower RHR is better
zSleep = (todaySleep - mean30d) / sd30d
z      = 0.5*zHRV + 0.3*zRHR + 0.2*zSleep  // renormalised if a signal is missing
score  = round(clamp(50 + 15*z, 0, 100))

Fourteen days of baseline minimum, thirty preferred. A standard deviation computed from four days is mostly noise, and dividing by it produces confident-looking nonsense. Below fourteen days Base reports "building baseline" with a countdown rather than a number. How the score is used.

What moves it that is not training

This is the list that stops you over-reading a single morning:

• Alcohol. Reliably and substantially, even in modest amounts.
• A late meal, particularly a large one.
• Room temperature, warm being worse.
• Illness onset, often a day before symptoms.
• Psychological stress, which the autonomic system does not distinguish from physical stress.
• Travel and altitude.
• Dehydration.
• Measurement inconsistency — a different device, a different time, a different posture.

That last one deserves emphasis. HRV measured on waking, lying down, before caffeine, is a different quantity from HRV measured after standing up. Overnight averages from a wearable avoid this by construction, which is one of their advantages over a spot morning reading.

Reading it as a hybrid athlete

The property that makes HRV useful here is that it is systemic and discipline-blind.

It does not know whether the fatigue came from a threshold session or from three heavy lower-body days. For a hybrid athlete that is exactly right, because the actual problem is almost always that the two loads combined exceeded what either alone would have — and a lifting-only fatigue model or an endurance-only one will both miss it.

It also has a specific limitation worth naming: HRV responds to metabolic and autonomic stress more than to mechanical stress. A heavy squat session produces a large amount of muscle damage and neural fatigue and a relatively modest HRV response. So HRV under-reads lifting load in the same direction that a cardiovascular strain metric does.

That is why it is 50% of a recovery score rather than the whole thing, and why the deload trigger also reads RIR drift — which is a mechanical fatigue signal that HRV structurally cannot see.

What to actually do with it

Do not act on one reading. One low morning after a wedding is a wedding.

Act on a trend of seven to ten days alongside a second signal. Two independent fatigue signals firing at once is the bar: recovery depressed against your baseline, RIR drifting up at fixed load, three or more short nights, or a flat strength trend.

Do not skip training on a low reading. Train slightly further from failure instead — one extra rep of RIR on today's targets. Skipping is almost always the wrong response to a moderately low score, because it costs the stimulus and the habit and returns very little.

Expect it to fall after hard sessions. That is the system working. A block with no HRV dips is a block with no hard sessions in it.

What HRV cannot tell you

• Whether you are ill. It moves before symptoms; it does not diagnose.
• How much mechanical fatigue you carry. See above.
• How you compare to other athletes. Never.
• What your training should be. It is one input to a decision, not the decision.

Base is a training log, not a medical device. A persistently depressed HRV with symptoms is a doctor's question.

What to do next

Check whether you have thirty days of continuous data. If your device comes off to charge or you do not wear it overnight, your baseline has holes in it and every reading against that baseline inherits them — which is a far more common problem than any question about which statistic to use.