One night of good HRV, a week that said otherwise

Last night's overnight HRV read 83 ms, 3.6% above the 7-day average of 80.1 ms. Today was marked on the calendar as a rest day, with nothing planned.

Taken alone, that reading looks like a good morning. Across the last 7 nights, though, HRV averaged 80.3 ms, down from 90.4 ms in the 7 nights before, a drop of 10.1 ms, or 11.2%.

What the two-week window shows

Over the full 14-night window, the mean is 85.4 ms, with a spread, or standard deviation, of 9.9 ms. Subtract the mean from last night's 83 ms and the difference is -2.4; divide that by the spread and the z-score is -0.24, a small distance either way.

A resting heart rate of 52 bpm told a slightly different story. The 14-night mean is 49.2 bpm, so last night ran 2.8 bpm above it, a z-score of 1.16 and the third highest reading of the 14 nights. Total sleep was 9.4 hours, with 0.1 hours of it recorded as deep sleep and 1.4 hours as REM. Body battery at wake read 93.

None of these numbers, by themselves, settles what kind of night it was. A good HRV reading and a high body battery sat beside an elevated resting heart rate and very little recorded deep sleep. Reading any one of them alone would have told a simpler, and less complete, story than the full set does together.

Why a week beats one morning

Several studies have tested whether daily HRV readings can guide how hard to train. In a 2007 study of 26 healthy, moderately fit men, the group whose training was adjusted by daily HRV raised VO2peak from 56 to 60 ml/kg/min over 4 weeks, while a group following a fixed plan did not change significantly; the two groups' changes in VO2peak were not significantly different from each other¹. The HRV group's threshold for easing off was never a single morning's number: training eased only when HRV fell below a value set from the runner's own 10-day mean and spread, or trended down for two days running¹.

An 8-week study of 40 recreational runners found the HRV-guided group ran fewer moderate- and high-intensity sessions than a fixed-plan group, 13.2 against 17.7, and still improved a 3000-meter time trial by 2.1%, against a non-significant 1.1% in the fixed-plan group. The authors describe the difference between the two groups as small².

A review pooling 8 such studies and 199 people found HRV-guided training better for vagal-related HRV measures, but not significantly better for resting heart rate, maximal aerobic capacity, or endurance performance. The authors add that how to set an individual's own HRV baseline is still being studied³.

A further meta-analysis across trained endurance athletes put the overall effect of HRV-guided training on VO2max at a small 0.402, with better results in amateur and female subgroups⁴. Across these four papers, the pattern repeats: a measurable edge sometimes appears, but it is small, and more than one comparison between groups did not clear the bar for statistical significance.

A small effect across trials does not mean no effect for any one person. It means that, averaged across many athletes and many weeks, the difference between training by a fixed plan and training by HRV has so far been modest and not always statistically separable from chance³⁴. That is a different claim from saying HRV-guided decisions do nothing; it is a caution against expecting a dramatic shift from any single HRV-informed choice.

What this means for a rest day

Today carried no planned session; the calendar listed it as a rest day. The brief does not say why, only that nothing was scheduled. What it does show is that last night's single HRV reading, on its own, would not have flagged the week's 11.2% decline. Only the two 7-night averages, set side by side, showed the fuller shape.

• Write down a 7-night HRV average alongside today's reading.
• Compare that average with the 7 nights before it, the way the window above does.
• See one HRV reading as one point inside a 14-night picture.

What does a 7-night HRV average show that one morning's reading does not?

A single HRV reading can move for reasons unrelated to training: a late night, or simply ordinary noise between nights. The 2007 study above used a 10-day baseline for exactly this reason, so training decisions were not made on one number¹.

Does a weak link between HRV and sleep score mean one of them is wrong?

No. An r of 0.11 only says the two numbers do not move together closely in this window; it says nothing about which, if either, is more accurate on a given night.

What should someone do on a day like this?

Keep recording each night's numbers, then set the 7-night average on either side of today beside each other, the comparison used throughout this piece. A single metric read alone carries less information.