What my watch actually recorded, published in full

Everybody quotes their HRV. Almost nobody publishes the series it came from, which makes the number impossible to argue with and useless to learn from. So here is mine in full: every night of sleep, every Body Battery range, every readiness score and every verdict the watch handed down, straight out of the device database. It is 8 days of one 45-year-old on a rotating roster — not a study, and not enough data to conclude anything general. It grows by a week every week.

What makes this week worth publishing. Garmin spent four days calling my training Unproductive. On the last of those four days the same watch rated my readiness 87/100 and described me as well recovered. Two of its own metrics, flatly contradicting each other, on the same morning. The data below shows why both were right.

Sleep: when it happened, not just how much

Duration is the number everyone tracks and the least interesting one here. What matters on a rotating roster is where the sleep sits. Across these nights I went to bed between 19:18 and 22:15, and woke between 02:18 and 05:10.

Each bar is one night, placed on a real clock. Total 5.5–7.5 h, averaging 6.6 h.

Body Battery: the floor matters more than the ceiling

The number people screenshot is the morning peak. The number that predicts how a session goes is the daily minimum — how close to empty the day ran. Mine bottomed out at 5 on the worst day and never once recovered above 82.

Each bar spans that day's lowest to highest Body Battery. A bar that starts near the floor is a day that ran out, whatever the morning said.

Readiness and HRV: both climbing, slowly

These are two different measurements on two different scales, so they get two charts rather than one chart with two axes. Over the series readiness went 35 → 87 and the rolling HRV average went 26 → 31 ms. Both moving the right way, both slowly.

Training readiness

Out of 100, taken on waking.

HRV, rolling average

Milliseconds. The band Garmin calls balanced for me sits around 26–40.

And here is the contradiction

Training load is where it resolves. Acute load is roughly the last seven days; chronic load is the longer-run average that Garmin treats as fitness. Mine fell from 646 to 525 across the series, while acute load collapsed from 573 to 268.

Chronic load — the long-run average Acute load — roughly the last seven days

Both lines carry the same unit, so they share one axis. The bands underneath mark what the watch was calling my training status at the time.

So both readings were true. Day to day I was recovering: sleep improving, HRV climbing, readiness up to 87. Over weeks I was detraining: I had stopped putting work in, so the chronic line fell. Unproductive was not telling me I was tired. It was telling me I was fresh and doing nothing with it — which, after a week of backing off because the numbers looked bad, is exactly what had happened.

That is the trap this page exists to document. A bad week of recovery data makes you back off; backing off improves the recovery data; the improved data makes backing off feel correct. The way out is to read the two timescales separately, which no single score on a watch face can do for you.

Every number, nothing left out

Date	Sleep	In bed	Body Battery	Lowest HR	HRV	Readiness	Garmin's verdict
Sat 26 Sep	—	—	5–6	71	—	—	—
Sun 27 Sep	6.83 h	22:15–05:10	5–29	61	26 ms	35	Strained
Mon 28 Sep	5.88 h	20:36–02:30	17–50	54	26 ms	50	Strained
Tue 29 Sep	7.03 h	19:18–02:27	30–68	54	27 ms	69	Strained
Wed 30 Sep	5.47 h	20:20–02:18	8–82	56	29 ms	48	Unproductive
Thu 1 Oct	6.23 h	20:15–02:30	5–52	59	29 ms	58	Unproductive
Fri 2 Oct	7.53 h	21:04–04:36	24–64	57	30 ms	77	Unproductive
Sat 3 Oct	6.93 h	21:20–04:19	10–64	61	31 ms	87	Unproductive

Pulled straight from the watch's own database with GarminDB — no manual typing, no cherry-picking. The series gets one more week every week.

What this is not. One person, 8 days, one wrist. Body Battery and HRV status are Garmin's own proprietary scores, not clinical measurements, and the readiness number is a black box. Nothing here is advice and none of it generalises on its own. It is published because the alternative — quoting a single flattering number with no series behind it — is worse.

The decisions I actually make from these numbers are written out on the rules page, including the threshold I got wrong and had to move after it failed against this very data. The plan it produces is on the front page.