Beyond the guide · a first look

After the crash: the feeling heals faster than the body

Two of the questions the addendum used to only pose (does the body read differently after a crash, and does a crash leave a debt that carries forward) have now had a first pass through the data. They stay on the beyond layer, not the field guide, because they're newer and less settled than anything Wiggers' guide is graded on. But they're no longer just questions.

first tested · least-settled layer

The shape of getting back up Every crash aligned at its lowest day (t0), across two weeks · n = 29 crashes

Every one of the 29 crashes lined up at its lowest day, then averaged. The two curves tell different stories on the same days: the feeling snaps back in a sharp V, down to the nadir, then up to its usual 4–5 ceiling within about two or three days, while the overnight autonomic load stays raised around the crash and drifts down only slowly, over roughly two weeks. Two honesty notes carry the reading: the felt-state depth at the nadir is true by construction: a crash is defined as a run of low-felt-state days, so only its speed of return is informative; the overnight-stress curve is the independent, non-circular one, and its slow settle is the actual finding. Bands are the middle half (p25–p75) across 29 episodes, n=1, and the spread is wide. The two axes are separate, named scales; only each line's shape and timing carries meaning, not where they cross. From 2024 the stress level is citalopram-shifted, but this within-window shape is robust to that slow drift.

And the same channel tells a real crash apart from a passing bad day. Aligned the same way, the overnight stress around a dip (a single bad day) never lifts at all:

A real crash · n = 29

A transient dip · n = 79

Same axis, same scale on both. A crash carries an overnight-stress perturbation; a dip doesn't; the body registers one and not the other. So the watch's stress channel physically separates a real crash from a transient notch, which is the sharper half of why it's worth watching at all.

What the first test found

The body really is measurably different after a crash, but only on the autonomic channels. Compared against carefully matched deep-trough days that weren't crashes (the honest control for "any low day looks like this"), four of seven watch channels stay distinguishable for two or more of the first five days out: the overnight and all-day stress and body-battery signals. The resting heart rate and the felt-state don't: they just bounce back like any bad day. So there's a real autonomic recovery signature the feeling doesn't carry, which is exactly the split the curve above shows.

But the debt doesn't compound. The obvious worry (that recent crashes stack the odds of the next one) was tested directly and not supported: recent crash density didn't raise crash risk to any degree the data can stand behind (odds ratio 1.13, 95% CI 0.88–1.27, p = 0.17). So there's a lingering signature after a crash, but no measurable snowball. The body is different after a crash; that difference doesn't obviously make the next one more likely. And the mirror worry (that pushing too soon in that still-unsettled window tips you into a fresh crash) was put to its own dedicated pre-registered test, and came back the same way: no push-to-relapse signal the data can resolve.

This fast-feeling / slow-body ordering is the one the published literature would predict, not an anomaly: recovery after post-exertional crashes runs days to weeks (Moore 2023: ~12.7-day mean in ME/CFS), and the autonomic channels are the slowest of all to settle (Radin 2021: resting heart rate ~79 days versus steps ~32). A feeling that rebounds while the autonomic load lags is the expected shape. The honest limit: no published study paired a daily felt-state with an overnight autonomic index through the same crash, so the exact pairing here is inferred from neighbouring work.

How to hold this

Least-settled layer. The recovery-signature half is a descriptive characterisation, with no pass/fail bar by design; only the recovery-debt half is a tested "not supported."
n = 1, small. 29 crashes; the confidence intervals are wide and shown, not hidden.
The felt-state depth is definitional. Crashes are defined as low-felt-state days, so the informative half is the autonomic channels, never the feeling's dip itself.
Medication caveat. The stress and body-battery channels are citalopram-dose-shifted after 2024; the matched-control design and the within-window shape reduce, but don't erase, that.
Descriptive, no cause. "The body is measurably different after a crash," never "the crash caused a lasting change."

The full workings, in the research repo: the recovery curves (R9) ↗ · the recovery-signature assembly (R7) ↗ · HA-P6 result ↗ · HA-P7 result ↗ · the PEM-recovery literature review ↗

↑ Back to beyond the guide