Data synced through 2026-08-06. Sleep, HRV, stress, body battery and training readiness current; 736 activity sessions, the most recent on 4 August.
Last night
Bed 23:10, awake 06:36. Six hours thirty-nine minutes, the longest night in a week.
| Last night | 14-day avg | |
|---|---|---|
| Sleep | 6h39 | 6h05 |
| Deep | 67 min | 73 min |
| REM | 86 min | 59 min |
| Awake | 47 min | 18 min |
| Sleep score | 70 | 66.6 |
| Resting HR | 65 | 62.8 |
| HRV | 22 | 29.0 |
| Lowest SpO2 | 81 | 85.1 |
Last night was almost an exact inversion of the one before it. Two nights ago: REM collapsed to 11 minutes while HRV hit 37, near the top of the fortnight. Last night: REM recovered to 86 minutes and HRV fell to 22, the lowest reading of the last fourteen days.
Both of those cannot be a story about REM. Two nights with opposite sleep architecture produced opposite HRV readings in the wrong direction, which is a reasonable sign that night-to-night HRV here is noise around a mean rather than a response to anything.
Three numbers are less ambiguous.
Awake time was 47 minutes, against a fortnight average of 18 and a previous high of 36. That is the most fragmented night of the period, on a night that was otherwise long and structurally normal.
Lowest SpO2 was 81, the lowest of the fourteen days.
Resting heart rate was 65 for the second night running. The fortnight average is 62.8.
Against that, readiness came in at 74 with one minute of recovery time outstanding, and acute load has fallen to 53. By the watch’s own reckoning the body is rested. The heart rate does not agree.
The last two weeks
Comparing 24 July to 6 August against the fourteen days before it.
| Prev 14 | Last 14 | Change | |
|---|---|---|---|
| Sessions | 13 | 8 | -38% |
| Distance | 233.9 km | 134.8 km | -42% |
| Training load | 367 | 152 | -59% |
| Acute load | 203 | 125 | -38% |
| Steps/day | 12,538 | 8,323 | -34% |
| Sleep | 417 min | 365 min | -52 min |
| REM | 86 min | 59 min | -31% |
| Deep | 63 min | 73 min | +16% |
| Awake | 25 min | 18 min | -28% |
| Sleep score | 71.9 | 66.6 | -5.3 |
| Sleep stress | 25.9 | 29.0 | +12% |
| Body battery at wake | 59.6 | 52.9 | -6.7 |
| HRV | 28.2 | 29.0 | +0.8 |
| Readiness | 59.5 | 64.0 | +4.5 |
| Resting HR | 62.0 | 62.8 | +0.8 |
Two counts sharpen this more than any average does:
- Nights under six hours: 2, then 6. Tripled.
- Nights with resting HR at 64 or above: 2, then 6. Also tripled.
What this actually says
The training reduction is not marginal. Load is down 59 percent, distance 42 percent, and the last session was two days ago. Acute load has gone from a mid-July peak near 300 to 53 this morning. This is as clean a deload as the data contains.
Resting heart rate went up 0.8 bpm across it.
That is the finding. In a normal deload the resting rate falls, and it is usually the first thing to move. Here training stress was cut by more than half over two weeks and the number drifted the wrong way, while HRV improved by an amount indistinguishable from noise.
The sleep debt looks like the obvious candidate. Sleep fell over the same fortnight – an hour less per night, REM down a third, six of fourteen nights under six hours against two before.
It is the wrong answer, and the longer record says so plainly.
Sleep is not the driver
Widen the window from two weeks to fifteen months and the relationship inverts.
| Month | Overnight sleep | Resting HR | Nights under 6h |
|---|---|---|---|
| 2025-11 | 5.77h | 56.9 | 18/30 |
| 2026-02 | 5.58h | 54.6 | 17/28 |
| 2026-04 | 6.35h | 58.5 | 11/30 |
| 2026-06 | 6.50h | 63.6 | 9/29 |
| 2026-07 | 6.46h | 62.6 | 9/31 |
Sleep duration and resting heart rate move together, not in opposition. February was the worst-sleeping month of the fifteen – 5.58 hours a night, 17 of 28 nights under six – and it carried the lowest resting heart rate on record at 54.6. June was the best-sleeping month, and it carried the highest at 63.6.
Between February and July, sleep improved by roughly 53 minutes a night. Resting heart rate rose 8 bpm across the same span.
Whatever is raising the number, more sleep is not the answer to it, because more sleep already happened.
Two other candidates, both closed
Sleep architecture. Two consecutive nights produced opposite structures and opposite HRV readings in the wrong direction: REM of 11 minutes with HRV at 37, then REM of 86 with HRV at 22. Night-to-night HRV here is noise around a mean, not a response to sleep quality.
Breathing. The overnight SpO2 minima look alarming in isolation – frequently in the low 80s. But Garmin’s breathing-disruption record over 28 nights averages 1.5 disruptions per night, with a maximum of 7, across four to five measured hours each night. Low SpO2 readings unaccompanied by breathing events are what wrist pulse oximetry produces from motion and poor perfusion, not what airway obstruction produces.
What is actually left
Training load is down 59 percent and resting heart rate did not follow. Sleep has improved over five months and resting heart rate did not follow. Breathing is unremarkable. Sleep architecture is normal. Subjectively: rested and alert on waking, readiness 74, one minute of recovery time outstanding.
One number has drifted about 3 bpm since February and every other channel, objective and subjective, reports nothing wrong.
That is not a finding in search of an intervention. It is a single unexplained trend on an otherwise clean record, worth a mention at a routine appointment and not worth reorganising a life around. The useful conclusion of two weeks of data is a shorter list of candidates, not a longer one.