My Long COVID story, in data
I had four years of data before I ever found the guide.
For four years I wore a smartwatch and wrote down how each day felt, long before I'd heard of Laure Wiggers' Smartwatch Pacing guide, a free document many patients rely on. So when I found it, I had something rare: years of honest data about my own body, recorded before I knew what to look for. Wiggers asks readers to do one thing with her guide: learn what's normal for you, and decide for yourself whether her patterns are true for you. That is what I do here, with those four years. How much of what she describes actually shows up in my data? I do it for myself, and as a worked example of what anyone might learn from their own body and a smartwatch.
Of the crash tests, one held; the rest are honest negatives. The orthostatic patterns are a different axis, present in my body but not about crashes. A weather report, not an alarm.
This can't predict a crash. At best it's a weather report on the body: a rough read of the kind of week I'm in. It won't tell me what to do in the next hour. And it's a working read for one body, mine, which may not be yours.
Four years, as I lived them
The body in this story is mine. I’m Willem. I’ve had Long COVID since the spring of 2022, and a Garmin Forerunner 245 has been on the same wrist every day since August 2021, several months before I got ill. That’s the single most useful thing about this dataset: the same watch, the same wrist, the same person, recording from before the illness to now. The instrument never changed under me (no new watch, no algorithm migration), so there’s a clean pre-illness baseline to measure against.
Alongside the watch I’ve logged a daily felt-state score and, on the days that mattered, a written note about what was happening. My data isn’t the subject here; it’s the instrument every pacing claim gets tested against.
- 1,755 days tracked, one wrist
- 98.8% with Garmin biometric data
- 1,372 days with a felt-state score
- 686 days with a written note
- 29 crashes recorded
- 79 sub-threshold dips
The felt-state score runs a 1-to-10 scale, but in four years it never once rose above 6. Against a lived ceiling of six, a three is not a mild day; a crash, here, means two or more days at three or below.
Here is the whole of it: four years of how each day actually felt, scored by hand, with the crashes and the smaller dips running along the bottom: thinning over time, but never gone. This is the ground truth everything downstream is measured against.
| Chapter | Felt-state median range | Crashes | Transient bad days |
|---|---|---|---|
| pre-illness | — | 0 | 0 |
| high-crash years | 3.5–5.0 | 14 | 18 |
| stabler years | 4.0–5.0 | 13 | 29 |
| taper | 5.0–5.0 | 2 | 1 |
And the same four years seen a second way: through the watch, not through me. This is what my nervous system was doing overnight. Line them up and the honesty is the point: the felt-state above carries the arc; the watch carries texture.
The one real hardware limit: the Forerunner 245 is a 2019 device and records no heart-rate variability at all. The patterns that want HRV are tested through a proxy: the night-to-night variability of stress during sleep, which Garmin computes from the same underlying signal. One word to hold lightly throughout: what Garmin calls stress is a read of how activated the nervous system was, not a measure of mental stress: what “stress” really measures.
The full dated timeline
- Aug 2021 Baseline Garmin first worn; pre-illness baseline begins
- Apr 2022 Onset COVID infection; Long COVID onset
- Sep 2022 Data Daily gevoelscore logging begins
- Nov 2023 Crash Longest crash recorded: 14 days, lowest score 1
- Jan 2024 Intervention CPAP begins on a sleep-apnoea diagnosis later refuted; since stopped
- Apr 2024 Intervention Citalopram (SSRI) buildup begins
- Jun 2024 Intervention Citalopram reaches 30 mg plateau
- Mar 2026 Intervention Citalopram taper begins; an accidental natural experiment
- Today Now Still tracking. The pendulum is still settling.
What I found
Her guide reads two different things off the same watch, so what I found comes in two parts. “Inconclusive” is an honest state here, not a gap.
What I looked for, and what I saw
Most of her guide describes patterns to recognise, not predictions to test. Here is what I saw of the ones I read that way, the circulation side (POTS) especially. A crash verdict would be the wrong question for these; the full statement-by-statement accounting is in the workings.
- Walls of orange Stress that stays high when it should fall Wiggers C4 · H02b Descriptive · both axes how & why →
A read of why stress stays stuck: PEM, POTS, food, emotion, illness. Not a crash alarm.
- The U-dip A within-day stress dip Wiggers ties to blood volume and treats with electrolytes Wiggers C4 family · HA11 Orthostatic · present how & why →
Wiggers' circulation pattern, present in my data and time-varying (and unmanaged, I never acted on it). Not a crash predictor, and that is the point.
- Stress while sitting still Rest-stress with a motion filter Wiggers C4b · HA-C4b My extension · Beyond the guide how & why →
Rest-stress with my own motion filter. The crash test landed underpowered; it continues on Beyond the guide.
The crash tests
A smaller set of her patterns make an actual prediction: this warns of a crash. Those I could put to a locked, pre-registered test and judge with a verdict. Of the four, one held.
- The body that looks perfectly recovered Overnight recovery, and the parasympathetic swing Wiggers D2/D5/B4 · HA07d Partly how & why →
- The HRV signal we had to improvise Heart-rate-variability decline (via a sleep-stress proxy) Wiggers B1-B5 · HA07d Showed up how & why →
- Morning heart rate above baseline The classic Workwell 'RHR + 15' rule Wiggers A1/A3 · H01 Not found how & why →
- Too much, in the days before Heavy exertion in the 4-day lead-up Wiggers E1 · HA01b Not found how & why →
One more thing she names: medication. Wiggers warns that starting a medication can shift your heart rate, stress, and HRV, so you may have to relearn what’s normal for you. One of my four years ran on an SSRI at a changing dose, and the watch’s stress and body-battery numbers moved with it, by a fixed, measurable amount. A worked example of exactly what she flags. How I measured it, and subtracted it from the rest →
Going deeper
The deeper you go, the closer to the data and the further from advice. The story above is the whole of it for most readers; these go further.
- The field guide → Wiggers' guide held against four years of my data: the patterns I can recognize, the predictions I could test, and the parts my data can't reach.
- Beyond the guide → Patterns the data showed that her guide doesn’t name, like whether the body reads differently after a crash.
- The workings → For the curious and the skeptical: the method, the pre-registered tests, and the honest limits, or take the fifteen-minute tour.
- The investigation storyline → How the thinking actually went, act by act: the dead ends and rescues included.
- What else could explain it → The rival explanations: the treatments and artifacts that ran through the same four years, and where each one stands.
The person behind the data
I’m Willem. I’ve had Long COVID since the spring of 2022, and the pacing practice this site keeps referring to wasn’t handed to me; I worked it out by trial and error, with an occupational therapist’s help, over the years the data covers. The story of those years as data is already here: four years, four chapters for the shape of it, the recovery in six phases for how it actually went, and what a 3 actually means for the days themselves, in my own words. What the watch could never see (what those years were like to live) is a longer piece I still owe this site.
How I know what I know
Every finding on this site rests on a pre-registered test: the question, the data slice, and the bar for “this counts” were all locked before the test ran, and each was audited in a fresh session before being run. This is the discipline against fooling yourself: the thing that makes one patient looking at their own data trustworthy rather than just suggestive. The full apparatus is on this site, room by room, in the workings, or let it walk you through itself: the skeptic’s tour, six rooms, about fifteen minutes.
The documentation
The pipeline scripts, the methodology, the pre-registrations, the audit reports, and the closed-hypothesis registry are all open at github.com/wmasman/gevoelscore-app. Not the raw data (the felt-state numbers, notes, and calendar entries stay private), but everything around it.
What I read
The guide, the studies, and the clinical pacing rules behind the findings: from the Workwell Foundation and the Bateman Horne Center to recent Long COVID autonomic research, and the peer-reviewed work validating what a Garmin actually measures. The full list, grouped by topic with plain summaries, is its own page: reading & sources.
Get in touch
This is one careful worked example, openly offered. If you’re a patient, a researcher, or just curious, you’re welcome to reach out: the simplest route is an issue on the open research repository; it keeps the conversation public and findable for the next reader with the same question.