Heart Rate Recovery

This is Attemorvis's first in-house experiment — and we wanted to kick things off with something simple, physical, and genuinely interesting. Heart rate recovery (HRR) measures how quickly your heart rate drops after intense exercise, and it's one of the most accessible markers of cardiovascular fitness. The faster your heart rate falls, the more efficiently your body is recovering. Rather than just reading about it, we decided to test it ourselves.

Testing method

Two subjects — Max and Avery — completed a repeated sprint protocol at Trinity's football field. Each round started with a sprint of 25 yards back and forth twice (100 yards total), pushing heart rate into at least Zone 4. Once finished, the subject stopped at the start line, held one of three recovery positions, and remained there until their heart rate dropped back below 120 bpm. Each position was tested twice per person, giving us six peaks of data per subject to work with.

The three recovery positions tested were:

• Standing upright

• Hands on knees

• Kneeling

These were chosen because they're the positions you'd realistically find yourself in after a hard effort — whether on a field, a trail, or a gym floor. Heart rate data was collected continuously via Strava GPX output and exported to Excel for cleaning and analysis.

Into the data

With six recovery peaks per subject and three positions to compare, there's a solid dataset to dig into. The analysis looks at how quickly each position brought heart rate down, whether recovery curves behave linearly, and how the two subjects differed — POTS and all.

Important caveats before reading the results

Sample size. Two subjects, two trials per position. No result here is statistically significant in the traditional sense — the tests confirm this (all p > 0.05). The analyses describe patterns in the data, not proven facts.

Avery has POTS (Postural Orthostatic Tachycardia Syndrome), a condition where the autonomic nervous system abnormally elevates heart rate in response to upright posture. This directly confounds her standing recovery data — her HR stays elevated when standing not because standing is a bad recovery position, but because her dysautonomia is actively working against her. Her data is also interpolated from a WHOOP device rather than raw 1-Hz sensor data. For these reasons, Max's data is used as the primary source for all mechanistic analyses. Avery's data is included where noted for comparison.

Summary of findings

Across all five analyses, kneeling produced the fastest and deepest heart rate recovery for Max, followed by hands on knees, with standing the slowest. The difference is most visible in the first 60–100 seconds after peak HR, where kneeling triggers a much steeper immediate drop. Both positions that reduce the hydrostatic blood column (kneeling, hands on knees) outperform standing upright, which is consistent with established sports-science literature on post-exercise vagal reactivation.

Kneeling produced the largest drop at every early checkpoint. By 30 seconds, Max's heart rate had already fallen 39.5 bpm in the kneeling position — more than double the standing drop of 17.5 bpm. By 60 seconds kneeling was ahead by 23 bpm. By 120 seconds, kneeling and hands on knees were both well clear of standing (65.5 and 56.0 vs 33.5 bpm).

The 180-second numbers are slightly less clear because Max's recovery windows ended before 180s in some kneeling trials (next exercise bout started), so those numbers are truncated.

Avery's results (confounded by POTS)

Avery's standing trial 1 shows an extreme HR60 of 97 bpm — this is an artefact of a very short 60-second recovery window before the next exercise bout, not a genuine recovery rate. Her other trials show kneeling and hands on knees performing similarly to Max's results. Her standing trial 2 (27.5 bpm at 60s) is lower than Max's standing (46 bpm at 60s), consistent with POTS slowing her standing recovery.

Key finding: Kneeling and hands on knees show the bi-exponential structure in 3 out of 4 trials, meaning these positions produced a distinct two-phase recovery with a clear fast drop early on. Standing's recovery is better described by a simple exponential or even a linear model — suggesting it lacks that sharp initial parasympathetic rebound.

The bi-exponential also fits kneeling T2 with the lowest AIC of any trial (567.9), meaning the data there matched the two-phase model particularly cleanly.

The linear model consistently underperforms (lower R², higher AIC than both curve models), confirming that HR recovery is not a straight-line process.

Key finding: Kneeling's early slope of −49.2 bpm/min is by far the steepest of any position — nearly double hands on knees (−27.3) and more than triple standing (−15.1). This means kneeling triggers the most powerful immediate parasympathetic reactivation in the first ~98 seconds after peak HR.

The slope ratio (early ÷ late) tells you how pronounced the two-phase structure is. Hands on knees has the highest ratio (8.8×), meaning the transition from fast to slow phase is particularly sharp. Kneeling has a very steep early phase but also a stronger late phase than hands on knees, suggesting it sustains recovery longer.

Standing's early slope (−15.1 bpm/min) is less than a third of kneeling's, and its breakpoint arrives later (119.5s vs 98.5s), meaning the initial recovery is slower and takes longer to even reach the transition point.

A notable individual result: kneeling T2 had an early slope of −75.6 bpm/min — the steepest single recovery ramp observed in the entire dataset.

With only 2 trials per position, traditional confidence intervals are unreliable. This analysis uses block bootstrapping — resampling the 1-Hz time series data in 10-second blocks (to preserve the autocorrelation in the signal) 2,000 times — to estimate the uncertainty around each position's mean recovery slope. 

Key finding: The confidence intervals for standing and kneeling do not overlap — standing's upper bound (−5.1) is well below kneeling's lower bound (−7.0). This suggests the difference between these two positions is real, not noise, even accounting for the small sample.

Hands on knees occupies the middle ground with overlapping CIs relative to both other positions, consistent with it being genuinely intermediate in recovery speed.

Kneeling also shows the widest CI (range of ~8 bpm/min), reflecting more variability between its two trials. This variability is partly because kneeling T2 was exceptionally fast (−11.3 bpm/min individual slope) while T1 was more moderate (−9.3 bpm/min).

The violin plots in the figure show the full bootstrap distribution for each position — kneeling's distribution sits visibly lower (faster recovery) and is wider than the others.

Key finding: This is the most nuanced result in the dataset. Standing reaches its plateau fastest (130s avg), which sounds good — but it plateaus at a much higher HR (132.5 bpm), having dropped only 40 bpm from peak. In other words, standing stops recovering sooner because it has already stalled at an elevated HR, not because recovery is complete.

Kneeling takes longer to plateau (214s) but reaches a significantly lower resting HR (119.0 bpm) with a total drop of 64 bpm — 60% more than standing's drop. Hands on knees is again intermediate (169s, 119.5 bpm, 61.5 bpm drop).

The practical interpretation: if you want your heart rate back down to a low number as quickly as possible, kneeling and hands on knees are better choices. If you stop caring about HR recovery after ~2 minutes (because you need to do something else), standing gets you to a stable state faster — just not as low a state.

Overall Conclusion

Every analysis points in the same direction for Max's data: 

Kneeling is the best recovery position in this dataset. It triggers the fastest initial parasympathetic reactivation, achieves the deepest total HR drop, and shows the cleanest two-phase recovery structure.

Hands on knees is consistently second, and meaningfully better than standing on every metric except time-to-plateau (where its advantage over standing is due to reaching a lower HR, not stalling early).

Standing performs worst across all metrics. It reaches a stable HR fastest, but that stable HR is substantially higher — the autonomic system appears to reach a compromise equilibrium rather than driving recovery as aggressively.

The physiological explanation is well-supported: both kneeling and hands on knees reduce the hydrostatic pressure column that blood must be pumped against, decreasing the cardiac afterload and allowing the parasympathetic nervous system to pull HR down more aggressively. Standing keeps the full column of blood below the heart, requiring sustained sympathetic output to maintain blood pressure, which directly competes with HR recovery.

For Avery, the pattern is similar for kneeling and hands on knees, but her standing data is confounded by POTS and should not be interpreted as representing a healthy standing recovery response.

Analysis scripts: analysis/ | Raw figures: figures/ | Raw data: data/