May 2026 ยท Data companion to Resistance Wearing a Lab Coat

Training Through The Trough

The H10, WHOOP, and Garmin receipts behind the recovery-score essay.

This is a narrow receipt, not a training plan. For ten consecutive days in April 2026, I had overlapping data from a Polar H10 chest strap, WHOOP, and Garmin while I trained through a recovery-score trough. The question was simple: did the score know enough to be obeyed?

The short version: after the Apr 15 hard run, WHOOP recovery dropped to 29 percent. The standard reading would have been rest. I kept training, but redirected the work toward steady aerobic sessions. Four non-green days later, the same walk came in about ten beats per minute lower than the comparable walk before the trough, with essentially the same cardiac drift.

The boundary: this does not prove red scores are wrong, or that anyone else should copy this week. It proves one narrower thing: for this body, in this window, a composite recovery score would have pushed against the training that appeared to produce the better later measurement.

The Timeline

The columns below are device outputs, not interchangeable scalars. WHOOP HRV, Garmin HRV, and H10 RMSSD use different collection windows and processing paths. The table is useful because it shows the sequence of events, not because the numbers should be merged into one master score. The same-walk table below uses session-level H10 values.

Date Stimulus WHOOP HRV Garmin HRV H10 RMSSD WHOOP Recovery
Apr 10 Rest 18.7 ms 32 - 48%
Apr 11 Hard mile, avg 144, max 175 26.3 ms 41 - 41%
Apr 12 Day after hard mile 32.7 ms 53 32.1 ms 88%
Apr 13 Rest - 36 21.1 ms -
Apr 14 Walk plus Zone 2 ride, avg 127 - - 26.7 ms -
Apr 15 Hard run, avg 153, max 182 25.4 ms - 23.0 ms 50%
Apr 16 Day after hard run 19.3 ms 46 17.4 ms 29%
Apr 17 Zone 2 to Zone 3 drift ride 19.1 ms 37 21.6 ms 31%
Apr 18 Zone 2 to Zone 3 ride, avg 142 21.6 ms 35 21.3 ms 51%
Apr 19 Ride, avg 139 22.5 ms 40 22.3 ms 61%
Apr 20 Walk, avg 103, best efficiency 20.8 ms 46 20.6 ms 46%

The two important moments are the peak and the trough. The Apr 11 hard mile produced the best recovery reading in the window the next day. The Apr 15 hard run produced the expected dip. If the device was treated as a command layer, Apr 16 and Apr 17 were rest days. I did not treat them that way.

The Same-Walk Comparison

This is the row that carries the argument. Apr 20 was not a maximal performance. It was an ordinary walk. That is why it matters. It shows lower cardiac cost on a comparable easy route after training through the low-score window.

Session Date Duration Avg HR Max HR RMSSD DFA-alpha1 Drift
Mar 26 walk 2026-03-26 94m 87 118 16.6 1.576 +15.9%
Apr 14 walk 2026-04-14 41m 113 127 5.1 1.566 +2.3%
Apr 20 walk 2026-04-20 54m 103 121 8.0 1.589 +2.2%

The Mar 26 row is in the table for honesty, not for the argument. It was a ninety-four-minute afternoon walk, about twice as long and easier than the others. Cardiac drift climbs with time on feet, so its higher number is a duration artifact, not a fitness signal. The matched comparison is the two walks of similar length and effort: Apr 14 before the trough, Apr 20 after it.

Apr 20 had lower average and max heart rate than Apr 14, better RMSSD, and nearly identical drift. The H10 quality check was clean: 420,480 ECG samples, 5,547 RR intervals, and 18 rejected artifacts, or 0.3 percent.

What The Wrist Devices Missed

The point is not that wrist devices are useless. The point is that the measurement layer is lossy, and the recommendation layer usually hides that loss from the user.

Apr 20 walk H10 chest strap WHOOP wrist Garmin wrist
Avg HR 103 100 No activity detected
Max HR 121 118 No activity detected
Duration 54.1 min 36.3 min No activity detected
Capture notes Full session ECG Captured 67% of the session Daily summary never caught the walk

What This Proves

Supported

  • In this ten-day window, the recovery score would have discouraged training during a sequence that later showed better cardiac efficiency.
  • The Apr 14 and Apr 20 same-walk comparison is a clean personal receipt for lower cardiac cost at comparable easy effort.
  • Wrist-device summaries missed or disagreed on details that mattered to interpretation.
  • Subjective state and session context were necessary to make sense of the numbers.

Not supported

  • That everyone should train through red recovery scores.
  • That low HRV is good, high HRV is bad, or any single direction is universally meaningful.
  • That WHOOP or Garmin are useless.
  • That a ten-day personal sequence establishes a population rule.

Data Provenance

Source Captured by Range
H10 walks Polar H10 chest strap, local NDJSON event stream Mar 26, Apr 12, Apr 14, Apr 20
H10 overnight Polar H10 chest strap Apr 19 to Apr 20
WHOOP HRV, recovery, sleep WHOOP wrist device, native app output Apr 10 to Apr 20
Garmin HRV, sleep, activity Garmin wrist device, Garmin Connect output Apr 10 to Apr 20

This page is the data behind the essay. The essay makes the argument. This companion keeps the receipts and the boundaries visible.