What a Normal HRV Baseline Looks Like for a Runner

Search for a normal hrv range for runners and you will get numbers between 20ms and 200ms, usually with no mention of which HRV metric produced them, which device measured them, or what time of night the reading was taken. That range is technically true and practically useless. A 34-year-old sub-3 marathoner can sit at 45ms and be perfectly healthy. A 52-year-old running four times a week can sit at 95ms. Neither number tells you anything until you know what that person’s own average looks like over a month.

So this page does two things. It tells you what the population numbers actually are, with the caveats attached, so you can sanity-check that your watch isn’t broken. Then it walks through how to build the only baseline that matters for judging an AI-generated plan: yours, measured the same way every night, expressed as a band rather than a single figure.

The population numbers, with the fine print attached

Most consumer wearables report overnight RMSSD (root mean square of successive differences) in milliseconds. Here is roughly where healthy, recreationally active people land, drawn from the distributions these devices publish and from the HRV literature they’re built on:

Three things to hold onto. First, the spread within any one age row is wider than the difference between rows, which is why comparing your number to a friend’s is close to meaningless. Second, HRV declines with age on average by something like 0.5–1% per year, so a 45-year-old whose number sits below a 25-year-old’s is not detraining. Third, the endurance column is a tendency, not a rule: plenty of well-trained runners sit in the middle of the general range, and there is no evidence that a higher absolute HRV predicts a faster 10K.

If your reading falls outside those ranges entirely, suspect the measurement before you suspect your heart. Optical wrist sensors on a loose strap, a night of alcohol, a late meal, or sleeping in a hot room will all move the number. Genuinely low readings with a genuinely well-worn strap, sustained over weeks, are worth a GP conversation rather than a training tweak.

Which HRV number your device is actually showing you

This is the step most runners skip, and it causes more confusion than anything else. “HRV” on two different apps can mean two different calculations over two different windows.

• Garmin (Forerunner 255/265/955/965, Fenix 7/8, Venu 3) records RMSSD continuously through the night and reports a seven-night rolling average, plus an “HRV Status” of Balanced, Unbalanced, Low or Poor. It needs about three weeks of nights with the watch on before it will show a status at all.
• Whoop takes RMSSD from your last slow-wave sleep period, so a single value from one slice of the night, compared against a weighted 30-day baseline.
• Oura reports average HRV across the whole sleep period and also shows the overnight trend line, which is genuinely useful for spotting a reading that spiked late.
• Polar feeds HRV from roughly the first four hours of sleep into Nightly Recharge, and reports an ANS Charge score from -10 to +10 rather than milliseconds.
• Apple Watch logs SDNN, not RMSSD, mostly during Breathe/Mindfulness sessions and at sporadic intervals. SDNN numbers run higher than RMSSD for the same person, often 40–120ms, and are not interchangeable.
• HRV4Training and Elite HRV use a one-minute guided morning reading, usually reported as lnRMSSD (the natural log) or as a 0–10 score.

Because of the window differences, Whoop and Garmin can differ by 15–20% on the same night on the same person, and both can be “right”. Pick one source, wear it consistently, and never mix numbers from two apps into one trend line. The broader question of how these scores get combined into readiness and recovery figures is covered in the parent guide on reading readiness, HRV and recovery scores, which is worth reading alongside this page if you’re also trying to interpret Body Battery, Training Readiness or a Whoop recovery percentage.

Building your own baseline: 30 nights, a mean and a spread

Your baseline is two numbers: a central value and a normal amount of night-to-night wobble. Thirty nights is enough to establish both. Here’s a real-shaped fortnight from a 38-year-old in week 5 of a 16-week marathon block, Garmin overnight RMSSD:

Mon  62   Tue  71   Wed  58   Thu  66   Fri  74   Sat  55   Sun  69
Mon  64   Tue  60   Wed  72   Thu  57   Fri  68   Sat  63   Sun  70

Mean across those 14 nights: 65ms. To get the spread, take the first seven nights (62, 71, 58, 66, 74, 55, 69), mean 65. The deviations are -3, +6, -7, +1, +9, -10, +4. Square them: 9, 36, 49, 1, 81, 100, 16, summing to 292. Divide by 6 (n-1) to get 48.7, square-root it, and the standard deviation is 7.0ms.

That gives a coefficient of variation of 7.0 ÷ 65 = 10.8%. For most runners the CV sits somewhere between 5% and 12%. Below about 5% often means your device is smoothing more than it’s measuring; above 15% usually means inconsistent sleep timing, alcohol, shift patterns, or a strap that’s moving.

Your working band is then roughly mean ± 1 SD, so 58–72ms here. A single night at 55 is inside normal noise. A single night at 44 is a genuine outlier worth noting. Seven-day average dropping to 56 is the signal that matters, because the rolling average filters out exactly the noise that makes single nights unreadable.

What Garmin’s “Balanced” range is doing under the hood

Garmin’s HRV Status screen displays something close to this:

HRV STATUS          Balanced
7-day average       65 ms
Your range          57 – 78 ms
Last night          62 ms

That “Your range” is built from roughly 90 days of your own nights, with the balanced band set around the middle of your personal distribution rather than any population figure. Two consequences follow. A brand new watch will show no status for three weeks and a slightly wrong one for six, because the band is still filling in. And if you slide into a chronically suppressed state slowly over two months, the band slides with you and eventually calls the new, lower normal “Balanced” again. Keep your own record of where the band sat in January if you want to notice that in April.

Two runners, same AI plan, different baselines

Consider two people running the same Runna 16-week marathon plan.

Runner A is 29, baseline 88ms, CV 9%, band 80–96. Week 9 brings a 22-mile long run. Tuesday’s reading is 71, Wednesday 74, Thursday 79. That’s three nights below band and then a return: normal, expected, no action.

Runner B is 47, baseline 41ms, CV 11%, band 36–46. Same week, same plan structure scaled to their paces. Their readings run 33, 31, 34, 32, 35 across five nights, with resting heart rate up from 48 to 54 and the seven-day HRV average down to 34. In absolute milliseconds, B’s dip is 8ms against A’s 17ms. As a proportion of baseline, B has dropped 17% and stayed there for five days while A dropped 14% and recovered in three. B is the one who needs the next quality session converted to easy running, not A.

The comparison that matters is percentage change from your own mean, sustained over days. Absolute millisecond differences between two people tell you nothing about who is more cooked.

What a normal block actually looks like

Expect your baseline to move over a training block, and expect the direction to be counterintuitive at times.

During a well-pitched build, many runners see the seven-day average hold steady or creep up by 3–8% over eight to twelve weeks as aerobic fitness improves. During peak volume weeks, a 5–10% dip in the rolling average that recovers within a down week is normal adaptation. A taper commonly produces a rise of 5–15% in the final two weeks, which is the pattern you actually want to see before a goal race.

The concerning shape is different: a rolling average that drifts down 10% or more and stays there across two to three weeks, usually with resting heart rate up 4–6bpm, easy paces feeling harder at the same effort, and sleep getting worse rather than better. One or two of those signals is noise. Four together, held for a fortnight, is the point at which an AI plan needs overriding rather than trusting.

Feeding this back into your plan

Runna and Garmin Coach build their schedules from your race goal and paces, not from last night’s HRV, so nothing happens automatically when your numbers slide. Garmin’s Daily Suggested Workouts do factor Training Readiness (which includes HRV Status), but only if you’re following those rather than a fixed coach plan. Check what your own app does before assuming it’s watching.

If you’re using ChatGPT or Claude to build or adjust a plan, give it the band, not the number. Something like: “38, week 9 of 16 to a marathon, Garmin overnight RMSSD baseline 65ms with a normal range of 57–78, seven-day average now 54 for nine days, resting HR 47 against usual 43, legs fine but easy pace 20s/km slower at the same heart rate.” That gets you a useful answer. “My HRV is 54, is that bad?” gets you a paragraph of generic reassurance, because the model has nothing to compare 54 against.

Write your baseline band and CV somewhere you’ll see them: the top of a training log, a note on your phone, the notes field of your Strava description if that’s where you live. Recalculate the band every six to eight weeks, and keep the old value written down next to the new one.