The most actionable finding is the HRV-to-volume relationship: nights when your HRV is higher, you tend to train more volume the following day (r=0.231), which suggests your body is already self-regulating — you push harder when you're recovered, whether you're consciously aware of it or not. Rather than trying to force volume on low-HRV days, lean into this pattern and treat low-HRV mornings as a signal to stay closer to prescribed weights rather than chasing rep PRs. The step-to-body-battery drain is the strongest signal in the dataset (r=-0.344 same day, still present at r=-0.181 a week later), meaning your general daily movement load is a real cumulative cost — on days before your heaviest sessions, keeping steps in the 9-12k range rather than 18k+ is probably worth prioritizing. The deep sleep finding is genuinely interesting: better deep sleep predicts higher steps seven days later (r=0.130), which likely reflects a recovery-to-capacity cycle rather than anything you'd directly manipulate, but it does reinforce that protecting sleep quality has downstream effects that show up nearly a week out. The main caveat is that these correlations are modest in magnitude — even the strongest (r=-0.344) explains only about 12% of the variance in body battery — so none of these are deterministic rules, and with 40 tests run, the FDR correction is doing real work to separate signal from noise, meaning the borderline findings (p around 0.08-0.10) should be treated as directional hints rather than firm conclusions.
Significant Correlations 8
Higher step days strongly predict lower end-of-day body battery.
Higher HRV nights precede higher training volume.
Higher step days strongly predict lower end-of-day body battery.
Higher HRV correlates with lower average stress.
More total sleep associates with lower resting heart rate.
More deep sleep associates with higher step counts.
More total sleep associates with lower resting heart rate.
Higher HRV correlates with lower average stress.
Methodology: Pearson correlations computed across 8 metric pairs at 5 lag windows (0, 1, 2, 3, 7 days), yielding up to 40 tests total. Multiple-comparison correction applied using Benjamini-Hochberg FDR at α = 0.10. Minimum 15 overlapping observations required per pair. Positive r (green) indicates metrics that rise and fall together; negative r (amber) indicates an inverse relationship.