Your heart doesn’t tick evenly.
That’s the good news.
At sixty beats per minute the gaps between beats are not an even thousand milliseconds. One is 980, the next 1040, then 995. That unevenness has a name — heart rate variability, HRV — and against intuition it is a sign of a system working, not failing. A metronome in your chest is bad news.
Accelerator and brake
Two branches of the autonomic nervous system steer the heart, pulling opposite ways. The sympathetic branch is the accelerator. The parasympathetic branch, through the vagus nerve, is the brake. A healthy system rides the brake constantly: speeding up on the in-breath, slowing on the out-breath, reacting to standing up, to coffee, to the thought of tomorrow’s race. HRV mostly measures the brake at work.
The number your watch shows is usually RMSSD — a measure of the differences between successive beats. The more correction, the more regulatory reserve. A rigid rhythm means the system is busy with something: an infection, short sleep, yesterday’s session, or stress that has nothing to do with running.
Three things the data show
| What | What it does to HRV | How much it weighs |
|---|---|---|
| Alcohol | Less brake, more accelerator. Recovery index drops by 9.3 / 24.0 / 39.2 points for low / moderate / high intake | Large sample, effect visible EVEN at low doses; being fit and young does not protect you |
| Incoming infection | The drop often precedes symptoms | Same signal shape as after drinking — a single night cannot tell them apart |
| Age | At 25: nocturnal RMSSD 80–120 ms. At 55 with moderate fitness: 25–45 ms | Both healthy. Comparing yourself to anyone else makes no sense |
One more number that surprises people: after a heavier evening the baseline returns not the next morning but after three to four days. For most of the week you are looking at data with Saturday still in it — and that tail is easy to mistake for overtraining.
Higher does not always mean better
This is the most interesting thing I found, and it contradicts what most apps imply. In runners driven into functional overreaching, weekly average HRV rose rather than fell. ↗ study The older model said the opposite: that a decline in parasympathetic indices heralds overtraining.
The practical consequence: the same rising curve can mean you are adapting beautifully or heading into overreaching. The number alone cannot separate them. Two-position measurement can: high supine HRV with a poor response to standing, plus declining performance, is the classic overreaching signature. In a 100 km walking ultramarathon study, slower finishers showed a LARGER drop in parasympathetic drive on changing position. ↗ study
Your number means nothing without your own band
If two healthy people can sit at 100 and at 30, the only sensible reference point is your own history. In UltraFlow I compute the baseline from a rolling sixty-day window and draw the norm band as the mean plus or minus one standard deviation.
And here is the detail that cost me a mistake: the mean and the band have different maturity thresholds. The mean is meaningful after fourteen nights. The band needs thirty — because with a shorter series a single artefact night can double its width. Showing someone a norm built from fourteen nights means drawing a boundary that will sit somewhere else tomorrow. The method that survives this was chosen by measurement across eight accounts, not by theory.
My system puts my baseline at 43.8 ms today. Runalyze, for the same person on the same day, says 39.86.
I write this instead of pretending the problem does not exist, because it shows something more important than the discrepancy itself: the baseline depends on the window and the method. As long as you stay with one source and one device, the trend is trustworthy. Comparing numbers across two systems or two watches is not. Validation studies say the same: wrist optical sensors systematically overestimate RMSSD, yet relative agreement is very high — good for tracking your own trend, useless for cross-device comparison. ↗ study
How much of this translates into training
Honestly: there is an advantage, but a modest and inconsistent one. HRV-guided training beat predefined plans on vagal indices, but not on resting heart rate. ↗ study For VO2max the effect was small and largest in amateurs and in women. Another meta-analysis found no difference at all. ↗ study And the most sobering observation is that HRV-guided groups simply did fewer high-intensity sessions — so the benefit may come from not overcooking yourself rather than from any magic in the metric.
That is why HRV never rules alone in my system. It is one signal beside resting heart rate, sleep and load — allowed to say “not today”, not allowed to rewrite your block. I wrote more about that boundary in the piece on how much of the mountains shows in the data; the foundations I calculate from are on the research page.
What to do with this tomorrow morning
1. Measure under the same conditions. Same time, same posture, before coffee and before your phone. HRV reacts to standing up more strongly than to yesterday’s session — measuring lying down one day and sitting the next compares two different things.
2. Look at the band, not the number. 42 means nothing. “42 against your band of 38–50” means “normal”; “42 against a band of 55–75” means “something is going on”. With fewer than thirty nights of history you don’t have a band yet — you have an average, and that is not enough to settle anything.
3. Never read a single night. A drop after alcohol, before an infection and after a hard long run all look identical, and the baseline takes three to four days to return. Base training decisions on the direction across several days and on how you feel — not on today’s reading.
Test it on your own data — 14 days free
Your own norm band, computed from your last sixty nights rather than a population average. A verdict that says plainly when the number fell outside the band and what to do about it today. 14 days free, card at signup, cancel any time.
Sources
Alcohol and nocturnal recovery, large sample — PMC · alcohol and nocturnal heart rate, smartwatch — PMC · parasympathetic hyperactivity in overreached runners — PMC · walking ultramarathon, supine and standing — PMC · HRV-guided training, methodological review — PMC · HRV-guided training and VO2max — PMC · device validation over 536 nights (2025) — Physiological Reports · wrist sensor overestimating RMSSD — PMC.