Heart rate variability declines with age. Not linearly, not uniformly, and with enormous individual variation — but the downward drift is real. Below are approximate overnight resting ranges most large-population studies converge on. Use them as rough orientation, not as a target.
RMSSD (Oura, Whoop, most overnight)
| Age | Low | Median | High |
|---|---|---|---|
| 20–29 | 35 ms | 60 ms | 100+ ms |
| 30–39 | 30 ms | 50 ms | 85 ms |
| 40–49 | 25 ms | 40 ms | 70 ms |
| 50–59 | 20 ms | 35 ms | 55 ms |
| 60–69 | 18 ms | 28 ms | 45 ms |
| 70+ | 15 ms | 22 ms | 35 ms |
SDNN (Apple Watch, ECG strips)
SDNN values run roughly 1.5–2× higher than RMSSD for the same person on the same night. An Apple Watch overnight SDNN of 50 ms at age 45 is roughly equivalent to an Oura RMSSD of 30 ms — both in the median band for that age.
Why these tables mislead more than they help
Population HRV distributions are wide. The 10th percentile and the 90th percentile in the same age band can differ by 3–4×. Genetics plays a large role. Body size matters. Resting heart rate matters. Sleep quality the night you measured matters more than your true baseline.
Two specific failure modes to watch for:
- The 25-year-old with HRV 35who panics because the table says “low.” They’re fine. Their baseline is in the 10th percentile of the band. They can move it with practice.
- The 55-year-old with HRV 60who feels validated because they’re “high.” They might be. But if their seven-day average drifted down 15 percent last month, that’s the signal — not the absolute number.
The actually useful comparison
Your seven-day rolling average versus your seven-day rolling average a month ago. That’s the comparison that catches meaningful change without drowning in single-day noise.
Decline of 15 percent: something has shifted — investigate. Increase of 15 percent: your interventions are working — keep going. Anything in the ±5 percent band: signal vs noise is ambiguous; give it another week.