Means across the selected period. Recovery and HRV climbed post-marathon while strain and energy burned fell — the classic shape of a deload.
Sleep stages by ISO week and the running sleep-debt line. Asleep duration is the only sleep variable that significantly predicts recovery (β=0.135, p<0.001).
Across the window, deep sleep held steady (~98 min) while sleep debt fell 13.7% post-marathon (77.5 → 69.4 min). The taper and post-race weeks (W18–W20) carried the lowest strain and the cleanest sleep of the year — debt bottomed near 42 min in late April before drifting back up as training resumed.
Does strain predict next-day recovery, and does it matter what kind of training? The honest answer: weakly, and no.
The quadratic (inverted-U) term is not significant in any period (p=0.20 overall) — there's no clean "optimal strain zone" in this data. A weak negative linear trend holds (r=−0.215, p=0.0095): higher-strain days are followed by slightly lower recovery. No training modality produced a statistically significant next-day difference (all p>0.05). HRV shows a consistent but non-significant gradient — Rest > Other > Strength > Running — hinting that running suppresses HRV more acutely than lifting.
Cohen's d for each logged behavior on HRV — the metric it shifted most. Green helps, red hurts. Bars past ±0.5 are practically meaningful.
| Behavior | n (yes/no) | HRV d | Recovery d | Deep d | Sleep d | Flag |
|---|
The nasal strip stands out — but backwards: it associates with a −0.74 d drop in HRV (p<0.001), almost certainly because Logan reaches for it on already-compromised nights (congestion, travel, drink). Reading in bed is the clearest positive lever — +0.51 d on sleep performance (p<0.001). Hydration and electrolytes trend helpful on recovery but don't clear significance.
Recovery, HRV, Strain, and Deep Sleep by ISO week. Click any week's bar to drop into its daily breakdown.
OLS regression of next-day recovery on sleep architecture and prior-day strain. Significant predictors highlighted. HRV gets its own row because it carries the strongest independent signal.
| Predictor | β | SE | t | p | Sig |
|---|
How this report was built and what it can — and can't — claim.
Post-marathon n=43 is modest — period-2 effect sizes (and the dose-response vertex shift) are descriptive, not prescriptive. The dose-response quadratic never reached significance, so the "optimal strain zone" is not claimed. Modality comparisons all fell short of significance and several post-marathon cells dropped below the n≥8 floor. Intervention behaviors are self-reported journal flags — associations only, not causal. Nutrition/meal-annotation correlation was out of scope this run and is omitted rather than stubbed. All charts are powered by data baked into this file; nothing is fetched at runtime.