HRV, Sleep and Training Readiness
How to read a wearable the way a coach does — trend over single day, context over absolute number.
- Trend beats single reads.Compare a rolling 7-day average against your own 30–60 day baseline. Never compare your number to anyone else's.
- Use RMSSD.It reflects vagal tone, is validated in the literature, and is what modern wearables compute under the hood.
- 10% below baseline for 3+ days = flag.One bad day is noise. Three consecutive days is the system telling you it's accumulating stress faster than it's clearing it.
- Combine HRV, sleep, mood, soreness.No single metric predicts next-day performance well. Four cheap inputs together do.
- Adjust the session, don't skip blindly.Cap intensity, swap VO2 work for Z2, reduce duration — in that order. Full rest only on red.
What HRV actually measures
Heart rate variability is the millisecond-level variation between successive heartbeats. A healthy, well-rested heart does notbeat like a metronome — parasympathetic tone (the “rest and digest” branch of the autonomic nervous system) is constantly nudging each beat slightly earlier or later than the last. More variability means more vagal control. Less variability means your sympathetic system is dominating — training load, psychological stress, alcohol, poor sleep, impending illness, over-reaching.
Two HRV metrics matter for athletes:
- RMSSD(root mean square of successive differences) — dominated by parasympathetic input. This is the metric used in virtually all sports-science HRV research from the last decade and the one your wearable exposes under the hood. It's robust, responds fast to stress, and works well on short (2–5 min) samples.
- SDNN (standard deviation of normal-to- normal intervals) — captures both parasympathetic and sympathetic input, needs 24-hour recordings to be meaningful, more common in clinical cardiology than sport. Safe to ignore for training decisions.
Gold-standard measurement is a medical-grade ECG. The practical gold standard for athletes is overnight nocturnal measurement from a good wearable — several hours of quiet, supine, post-absorptive data averaged into one value. Far less noisy than a 2-minute morning reading, and fully automated.
Trend vs single read
HRV is a noisy signal. Day-to-day swings of 10–20% are routine even in well-rested athletes. A bad night's sleep, a spicy meal, dehydration, a late workout — any of these can punch a hole in tonight's reading without telling you anything useful about your training state.
Every reliable HRV framework — from Kiviniemi and Plews onward — converges on the same rule: compare a 7-day rolling average to a personal 30-to-60 day baseline. The 7-day window smooths out random noise. The 30-day baseline adapts slowly enough to reflect genuine fitness state, fast enough to respond to a new training block.
The shape of that signal looks like this:
Typical two-week pattern. Days 1–10: normal scatter around baseline — ignore. Days 11–14: sustained drop below the −10% line — act.
The decision rule: three to four consecutive days of the 7-day average more than 10% below your baseline is a genuine signal. Fewer than that — noise. Don't rearrange your training around a single bad Tuesday.
Sleep metrics that matter
Wearables report a wardrobe of sleep metrics. Only four predict training readiness reliably. The rest are estimates with large error bars and should be read as trends, not absolutes.
Total sleep time
The most validated sleep metric on any consumer device, typically within 10–15 minutes of polysomnography. Target 7–9 hours. Dropping to 4–6 hours costs 5–12% of endurance output. Below 5 hours, downgrade the session.
Sleep consistency
The standard deviation of your bed- and wake-times across a week. High consistency (< 30 min variation) correlates with better HRV, lower resting heart rate, and improved glucose regulation — independently of total duration. Going to bed at 22:00 every night matters more than any given night reaching 8 hours.
Deep and REM percentage
Deep (slow-wave) sleep is when growth hormone pulses and muscle repair happens. REM consolidates motor learning and emotional regulation. Wearables approximate these stages from heart rate and movement patterns — directional accuracy is decent, absolute numbers less so. Track the trend: a sustained drop in deep-sleep percentage often precedes a drop in HRV by a few days.
Resting / nocturnal heart rate
A rise of 5–7 bpm in overnight resting heart rate that isn't explained by training load is the earliest detectable sign of something off — illness incubating, dehydration, severe sleep debt. Much less noisy than single-day HRV. Watch RHR and HRV together; they tell you the same story from two angles.
Wearable validity — who's right?
Consumer HRV accuracy has improved dramatically since 2023. A 2025 validation study compared modern wearables against medical-grade ECG and found most top devices hit 0.94–0.99 concordance for nocturnal RMSSD. Device choice matters less than picking one and sticking with it.
| Device | Nocturnal HRV accuracy | Known gotchas |
|---|---|---|
| Oura (Gen 3 / 4) | Exceptional. ~0.94–0.99 vs ECG. | Finger pulse + skin temperature is the cleanest consumer setup. Reading quality degrades if the ring is loose or worn on a finger with cold extremities. |
| Whoop 4.0 | Exceptional. ~0.94–0.99 vs ECG. | The “Recovery” score weights deep-sleep parasympathetic rebound heavily, so it can say “green” on nights with short total sleep but good deep-sleep density. Cross-check with total sleep time. |
| Garmin (Enhanced BBI) | Research-grade on newer models with Enhanced BBI (Fenix 7+, Epix Gen 2, Forerunner 255/265/955/965). | Older Garmins under-report high HRV and over-report low HRV — systematic bias. If your watch predates Enhanced BBI, don't use it for HRV tracking. |
| Apple Watch | Spot-check only. Not continuous nocturnal. | Samples HRV in 1-minute windows at irregular times. Not suitable for a rolling baseline without third-party apps that force scheduled readings. |
| Polar H10 + app | Medical-grade — chest strap gold standard. | Requires a 3–5 min supine reading every morning. Most accurate single measurement available to consumers, but compliance is harder than wearing a ring. |
Absolute HRV numbers don't transfer between devices. Pick one, build a 30-day baseline on it, and only compare within that baseline.
Combining signals into a readiness call
No single signal predicts next-day performance well enough to drive training decisions on its own. The literature converges on a four-input framework:
- HRV deviation — 7-day rolling average vs 30-day baseline.
- Last night's sleep duration.
- Two subjective ratings — mood/stress and muscle soreness, each 1–5.
- Days since last hard session (Z4/Z5 intervals or long run).
Score each input, sum, and apply a threshold. This is the algorithm behind the recovery readiness score. The adjustment rules below show what to do with the verdict.
Adjustment rules — what to actually do
The one decision most runners get wrong: they either skip training entirely or do it exactly as planned. Neither is usually right. The correct move is almost always a graded modification.
| HRV deviation (7-day vs baseline) | Sleep | Today's session |
|---|---|---|
| At or above baseline | 7+ h | Green. Proceed as planned. Primed for intervals, long run, tempo. |
| 1–5% below | 7+ h | Green with awareness. Proceed. Reassess mid-warm-up; abort intervals if legs feel genuinely bad. |
| 5–10% below | 5–7 h | Amber — cap intensity. Keep the session but cut VO2 work to threshold, threshold to tempo. Hold duration. |
| 10–20% below | Any | Amber — swap modality. Replace quality work with upper Z2 aerobic. Cut volume by ~30%. |
| >20% below, 3+ day streak | < 5 h | Red. Full rest or 30 min lower-Z2 active recovery. Eat. Sleep. Re-test tomorrow. |
When the HRV row and the sleep row disagree, defer to the more conservative signal.
Special cases — when the normal rules bend
Female cycle phases
Luteal-phase HRV suppression (roughly days 15–28) is a hormonal artefact, not a training flag. Progesterone raises core body temperature and sympathetic tone, pulling RMSSD down a few percent and nudging resting heart rate up. Build a phase-specific baseline by averaging readings across 2–3 cycles and compare within phase. A 10% drop from your luteal-phase baseline still matters; a 5% drop from the follicular-phase mean into luteal territory is expected.
Masters athletes (>50)
Absolute RMSSD naturally declines with age — a healthy 55-year-old can have a baseline 40–50% below a 25-year- old's. Percentage-deviation rules still apply, but clearance from hard sessions takes longer (48–72 h is typical for VO2 work instead of 24–48 h). Skew amber calls toward rest more aggressively than a younger athlete would.
Alcohol, jet lag, illness
All three crush HRV in ways that have nothing to do with training load. If you drank the night before, flew across two or more time zones in the last 48 h, or are fighting something off, the reading is measuring that — not your fitness. Don't let it drive training decisions in either direction.
RED-S / under-fueling
Chronic energy deficit (Relative Energy Deficiency in Sport) produces persistently suppressed HRV that does not respond to rest or deload. The fix is nutritional, not training-based. If your baseline is drifting down over months rather than weeks, or you're hitting missed menstrual cycles, weight loss, or recurrent bone stress injuries, stop training adjustments and talk to a sports-medicine clinician. The IOC 2023 consensus on REDs is the reference.
Common mistakes
- Chasing the number.Trying to “raise your HRV” through ice baths, breathwork, cold showers, red-light panels. HRV is a mirror, not a dashboard to optimise. Sleep more, drink less, train consistently — the number tracks.
- Training hard when the trend is red. The trend line exists to tell you something. Ignoring three consecutive days of sustained drop is how overreaching turns into overtraining syndrome, which takes months to resolve.
- Skipping a workout because of one bad reading. The opposite failure mode. Single-day noise is not a training signal. If the other three inputs (sleep, mood, soreness) are fine, run.
- Comparing your HRV to someone else's. RMSSD is as individual as height. Your training partner's 85 ms vs your 50 ms means nothing about either of your fitness states.
- Ignoring subjective feel when the device says green. Mood and soreness ratings predict next-day performance about as well as HRV in the research. When the body and the watch disagree, default to the body.
Bottom line
Readiness data is useful in direct proportion to how coach-like you are about reading it. Track RMSSD overnight on one device. Build a 30-day baseline. Compare a rolling 7-day average against it. Combine with sleep duration, mood, and soreness. Act on trends of three or more days, not single reads. Adjust the session — cap intensity, swap modality — before you skip it outright.
If you want the decision automated: the recovery readiness score implements this framework as a ten-second check. If you want the wider context on sleep, fueling, mobility and temperature work, the pillar is Recovery for Runners: 6 Rules for Free Speed.