I tracked my heart rate variability for 5 years. Then I changed one thing, or thought I did.

I’ve measured my heart rate variability every morning since around 2005, and more recently with HRV4training since April 2021. Nearly five years of data using a Polar H9 or H10 chest strap, and the HRV4Training app, first thing after waking. For most of that time, I’d lie down for 2 minutes to take the reading, mainly because I’m not a morning person. In February 2025, I switched to sitting for 1 minute.

That position change broke my dataset in half. Something changed, and HRV improved. However, the improvement coincided with other changes: I’d ramped up my strength training with HYROX-related activities, as I’m not getting any younger, and I added yoga in January 2025. This makes the 2025 data difficult to decisively interpret – I can’t cleanly separate the measurement method change from genuine physiological adaptation.

The four years preceding those changes tell a clearer story and add important context.

What HRV4Training’s Six Metrics Mean & Show

HRV4Training tracks six core metrics based on timing differences between heartbeats (RR or NN intervals). Here’s what each did from April 2021 through January 2025, before I changed anything:

• rMSSD (root mean square of successive differences): Primary HRV metric. Responds to training stress, sleep, and illness. Statistical analysis shows a very small downward trend – roughly 0.3ms per year – but daily variability (coefficient of variation ~25-29%) completely swamps this drift. Practically speaking, it oscillated around the same baseline for four years.
• HR: Resting heart rate provides context. Small upward trend of roughly 0.1 beats per minute per year. Bounced within a narrow range across all years.
• SDNN (standard deviation of NN intervals): Total variability, better for 24-hour recordings. No statistically significant trend. The tightest of all metrics – barely moved across the entire period. This is the one Apple uses.
• pNN50 (percentage of successive intervals differing by >50ms): Another measure of beat-to-beat variation. No statistically significant long-term trend. Mean values jumped in 2022, then declined back through 2023-2024.
• LF (low frequency power, 0.04-0.15 Hz): Mix of stress (sympathetic) and relaxation (parasympathetic) signals, difficult to interpret alone. No trend detected.
• HF (high frequency power, 0.15-0.4 Hz): Linked to breathing rhythm. No movement whatsoever for five straight years.

The pattern across all six showed minimal directional change. Nearly five years of high-volume endurance training – 10-15 hours per week of running, cycling, and swimming – and my HRV metrics showed only trivial statistical trends. The slope changes were detectable mathematically but practically irrelevant relative to normal day-to-day variation. But…

What Near-Stability Actually Means

My data and charts show I’m neither overtrained (declining HRV) nor undertrained (rising HRV) over the longer term. Even in the short term, my HRV drops after hard sessions but recovers quickly.

I exercise 10-15 hours per week. It’s periodised when goal-focused, with peaks reaching 500+ TSS, and recovery weeks dropping to 100 TSS. thus probably ruling out monotony as a factor.

Near-flat HRV over five years suggests equilibrium. Is that good?

The 2025 Confound – Two Things Change Together

In late 2024, I added a weekly, hard yoga session and then an additional strength training or HYROX-style work in early 2025 with the goal of increasing flexibility and mass – the latter to counter age-related muscle loss and to keep the performances up.

One month later, in February 2025, I changed my waking HRV measurement protocol from lying down for 2 minutes to sitting for 1 minute, allowing approximately 30 seconds for my heart rate to stabilise.

This creates a problem: Other than not remembering exact dates, I can’t separate the measurement artefact from the genuine physiological response to the new training stimulus.

What Changed Immediately (Position Effect)

When I switched positions in February 2025, the breathing-related metrics changed as follows, with the other metrics barely moving, except for a 10% increase in pNN50.

• HR +3.7%: Sitting makes your heart work harder as it fights gravity.
• HF +33%: HF tracks how much your heart rate changes with each breath. Sitting alters your breathing pattern because a different posture gives your diaphragm more room to move.
• LF -11%: When lying flat, your body adjusts blood flow in slow cycles. A different position gives a different regulation pattern.
• The weird bit: Usually, when you sit up, HF drops, and LF rises (your nervous system kicks in). Mine did the opposite, perhaps because the breathing changes swamped everything else.
• rMSSD: +2.8% (not breathing related)
• SDNN: -1.3% (not breathing related)
• pNN50: +9.6% (not breathing related)

What Continued to Change (Adaptation Effect)

If position change were the only factor, metrics should stabilise within weeks. They didn’t.

Here’s the post-position data by ‘month’:

rMSSD:

• Days 0-28: baseline
• Days 29-56: +11.0%
• Days 57+: +12.1%

HR:

• Days 0-28: baseline
• Days 29-56: -0.5%
• Days 57+: +2.4%

Both show sustained upward drift (p < 0.0001), not stabilisation.

The most likely explanation is an adaptation to the strength training. Bodies don’t adapt overnight – neuromuscular efficiency, movement patterns, and recovery capacity improve over extended periods. Research shows that strength training can improve HRV in endurance-only athletes by improving autonomic balance. Despite a good cardiovascular system, my baseline strength was relatively low, leaving room for adaptation.

But I can’t prove this. The position change and the training change occurred within weeks of each other.

Why Waking+Sitting HRV Readings Make Sense for Athletes

Most guidance says lie down or use overnight averages – the “purest” measure without gravity affecting your heart.

But for athletes, that purity might miss the point.

When you’re very fit and lying completely still, your body reaches maximum relaxation. This creates a problem if you’re already at maximum relaxation, as HRV cannot decrease further to signal training stress. Studies show that overtrained athletes showed HRV changes when sitting or standing, but not when lying down.

So, sitting makes your heart work against gravity. This mild challenge reveals whether your body is handling stress well – better for detecting training readiness than measuring general health.

Duration and Breathing Protocol Matter

I didn’t just change position – I also changed duration. Two minutes to one minute. This also matters.

Shorter recordings increase noise, particularly for frequency metrics like HF and LF. A 60-second window captures fewer breathing cycles and fewer low-frequency oscillations.

My measurement variability increased post-change:

• HR variability: +33%
• HF variability: +50%

Some comes from position, some likely from shorter duration. another pesky confoundment.

Unpaced breathing: I’ve never used paced breathing during measurements. Marco Altini and others argue against it for baseline morning measurements. Paced breathing artificially inflates HRV; it’s useful for biofeedback training, not for tracking baseline stress. Spontaneous breathing reflects your actual state.

My protocol follows this principle, but because breathing rate wasn’t controlled and likely varied between lying and sitting, another uncontrolled variable was introduced.

Practical Takeaways

There is one key takeaway for me

and several smaller ones.

• Near-stability indicates sustainable training. HRV remaining approximately flat year over year, while training is consistently high, suggests sustainable volume.
• Position depends on the goal. For athletes monitoring exercise readiness, sitting makes more sense based on research. For general health monitoring, lying works fine.
• Duration affects quality. Shorter recordings increase noise, especially in frequency metrics, but are more convenient.
• Trends require statistics. Small, significant trends might not be obvious if day-to-day variation swamps them.
• rMSSD and HR tell most of the story. Others add nuance but rarely change interpretation.
• Equipment variation matters. Multiple “equivalent” devices over the years may introduce discontinuities. Keep a record.

The Lesson

Changes to my 2025 training and HRV tell a complex story. Multiple plotlines (strength training and position change) and a new character arc (duration change) are mixed and difficult to interpret because I changed too many elements at once. Clean data requires disciplined protocols.

One thing is certain: like you, I’ve got a bit older over the last five years. Probably by a similar amount to you, in fact. 😉

As an older athlete, I’ve reached the point where consistency matters. Perhaps even ‘slowing the decline’ would be a more appropriate phrase. Perhaps I’ve started winning races because, where others have given up or been injured, I’ve found a good, enjoyable and sustainable level at least for now.

Maybe (definitely) I could have tried harder and optimised more.

Last Updated on 15 February 2026 by the5krunner

the5krunner

tfk is the founder and author of the5krunner, an independent endurance sports technology publication. With 20 years of hands-on testing of GPS watches and wearables, and competing in triathlons at an international age-group level, tfk provides in-depth expert analysis of fitness technology for serious athletes and endurance sport competitors.