Heart Rate Variability (HRV) Biofeedback and Athletic Performance: Part Two

Important metrics

Marco Altini
7 min readMay 31, 2020

In this series of posts, I discuss Heart Rate Variability (HRV) Biofeedback in the context of athletic performance, and in particular:

If you are new to HRV Biofeedback, start from part one. Here, we’ll jump right into the most important metrics.

Heart Rate Variability analysis in the context of Biofeedback

In part one, I’ve explained how it seems that at least part of the positive effects of HRV Biofeedback can be explained by increased parasympathetic activity.

What metrics can be used to quantify this improvement?

When it comes to HRV Biofeedback, a standard HRV analysis of time and frequency domain HRV features is not necessary, as all that is required during an HRV Biofeedback session is to monitor how instantaneous heart rate varies over time, together with breathing rate, and use this information as feedback to help the user synchronize heart rate and breathing rate. Hence, we could be doing HRV Biofeedback sessions just with a heart rate monitor able to show real-time information (averaged heart rate over the last 15–30 seconds as provided by most gadgets is however not very useful, as we would lose the ability to see semi-real-time changes in heart rate due to breathing).

However, standard HRV analysis emerges as a simple non-invasive and quantitative instrument that could be used to quantify the effects of HRV Biofeedback. As a matter of fact, we will see how many of the interventions covered later in this series of posts do indeed report standard HRV features to assess physiological or psychological improvements.

Thus, in this section, I will first provide a short overview of HRV analysis. Then, I will also cover key aspects linked to acute and baseline changes in HRV that are unfortunately often overlooked and can explain inconsistencies found in literature, as well as overly optimistic results often found when employing HRV Biofeedback techniques in various applications.


HRV can be determined according to various methods. The first step is always the same and requires to compute time differences between consecutive beats, also called RR intervals (the heartbeat is identified by the letter R of the QRS complex when analyzing electrocardiographic data).

Example of ECG data and detected beats, original data. RR intervals are the time differences between two beats. As we can see there are always small differences between consecutive beats, and these differences reflect ANS activity and modulation of heart rhythm in response to various stressors. RR intervals can be used to compute HRV features once enough data has been collected (typically at least one minute).

Once RR intervals have been collected, typically HRV analysis methods are used to compute time and frequency domain features. While a deeper analysis of methods for HRV analysis is beyond the scope of this post (see Shaffer and Ginsberg, 2017), the Root Mean Square of the Successive Differences (rMSSD) and High-Frequency power (HF) features are often used in the time and frequency domains as markers of parasympathetic activity. The vagus nerve, the main nerve of the parasympathetic nervous system, acts on receptors signaling nodes to modulate heart rate on a beat to beat basis while sympathetic activity has different pathways with slower signaling. Thus, beat to beat changes computed as rMSSD or HF reflect parasympathetic activity very well as vagal influence has very short latency (less than 1s, Nunan, Sandercock, and Brodie, 2010). This is why most papers report rMSSD or HF as relevant metrics. The Low-Frequency power (LF) is also often reported. In earlier research, LF was thought to be a feature representative of sympathetic activity. At present, most researchers agree that LF is in fact representative of a combination of sympathetic and parasympathetic activity, and highly dependent on breathing frequency. Therefore, LF seems to be less relevant in the context of baseline changes in HRV, but we will see how this becomes one of the most important features during the breathing exercise, as at that moment we are breathing within the LF band.

Beware of the difference between acute and baseline changes in HRV

An important distinction needs to be made between acute and baseline changes when analyzing HRV during or after an intervention. In particular, acute and baseline changes in physiology refer to two very different aspects of the HRV Biofeedback intervention. Acute changes are almost instantaneous changes in HRV due to a particular event, which could be a stressor (e.g. a physical stressor such as exercise or a psychological stressor such as a math test or STROOP test), or a deep breathing exercise.

During an HRV Biofeedback session for example, greater changes in heart rate are elicited as a result of the session itself, mainly because RSA is maximized while breathing at a low frequency. As a result, HRV will increase, and the breakdown into various frequency domain powers (e.g. LF and HF) will be highly dependent on the breathing frequency. Typically, HRV Biofeedback sessions where the participant is breathing at 0.1 Hz result in increased LF and reduced HF, simply because the LF band includes 0.1 Hz.

Note that this does not mean that this change is not important. On the contrary, the large oscillations due to breathing at the resonant frequency are most likely what causes the positive outcomes reported by countless studies. However, it should be made clear that baseline physiology might still be unaffected and needs to be measured differently, outside of the biofeedback session.

Unfortunately, most research studies do not clearly explain when HRV is measured (in the morning, during the night, before the protocol, during the intervention or after the intervention). However, such acute changes in HRV Biofeedback might be unrelated to any baseline changes in HRV as measured using standard protocols for physiological assessment of baseline stress (for example measurements taken first thing in the morning or during the night, you can find plenty of case studies here or at the link below).

It is important to make this distinction and to carefully analyze literature results as only baseline changes in physiology are reflective of any positive adaptation or improvement due to the HRV Biofeedback intervention.

On the other hand, acute changes as measured during the session are merely artifacts of the deep breathing exercise. While these aspects are paramount to determine the effectiveness of the HRV Biofeedback intervention in terms of physiological changes, we will see in the next posts that baseline measures are hardly ever reported in the scientific literature.

Additionally, data is often collected only during the HRV Biofeedback sessions, and therefore provides no information on the effectiveness of the intervention in terms of improving baseline parasympathetic activity. Hence, I would highly recommend combining your HRV Biofeedback sessions with morning measurements (or night measurements) of HRV, taken for example using HRV4Training or the Oura ring, so that you can effectively determine if the HRV Biofeedback sessions are having an impact on your baseline physiology.

Lastly, a note on coherence. Coherence is a made-up HRV feature that represents how well a person is breathing near the resonant frequency, and was introduced by one of the main HRV Biofeedback device manufacturers. Similarly to what I have just discussed, coherence can track HRV during the session, but does not provide any information on baseline or chronic changes in physiological stress level or baseline parasympathetic activity.

For clarity, I have collected data before, during and after an HRV Biofeedback session so that you can see the acute effect of the session on instantaneous heart rate and standard HRV features:

Example of an HRV Biofeedback session, original data collected by the author. In the top panel (a) RR intervals are shown. RR intervals are the inverse of instantaneous heart rate, and we can see how deep breathing at the resonant frequency elicits large oscillations in instantaneous heart rate. We can also see how oscillations are much smaller during regular breathing at rest, shown in a darker gray before and after the HRV Biofeedback session. In the panels below, different HRV features are shown. In particular, panel (b) shows LF, the frequency power in the band that includes the resonant frequency. As a result, we can see how there is a clear increase in LF during HRV Biofeedback. This is what is often reported as an increase in HRV due to HRVB, but we can see how this increase is acute and most likely entirely due to RSA. We can also see how the exercise does not affect LF at all as soon as regular breathing is resumed, as highlighted by the lower LF the last two minutes. In the two bottom plots we can see HF and rMSSD, two features representative of parasympathetic activity. I have plotted both to show how HF is highly dependent on breathing frequency, and despite the great oscillations in RR intervals, the fact that we are breathing in a band outside HF results in quite inconsistent results, sometimes with values below resting values. On the other hand, rMSSD reflects parasympathetic activity with less dependency on the exact breathing frequency, and therefore captures better the increase in HRV due to the HRVB exercise. In most studies in literature, only LF or HF during HRV Biofeedback are reported, with a few studies providing baseline values similarly to what I have shown here in darker gray. Ideally, additional measurements should be taken first thing in the morning in order to assess baseline HRV for an individual outside of the HRVB session

HRV Biofeedback can facilitate self-awareness as individuals can see in real-time changes in physiological variables that can be influenced by their behavior and psychological state.

While different pathways have been proposed, and the mechanisms of HRV Biofeedbackare not entirely clear yet, the potential positive effects of HRV Biofeedback could be explained by increased parasympathetic activity (from a physiological point of view) and how this links to neural activity in terms of reduced anxiety for example (from a psychological point of view). All of these factors can potentially impact performance positively.

In the next post, I will provide an overview of the protocols used in HRV Biofeedback interventions to improve emotional self- regulation and performance, relying on the theoretical and physiological framework covered in parts one and two of this guide.

Marco holds a PhD cum laude in applied machine learning, a M.Sc. cum laude in computer science engineering, and a M.Sc. in human movement sciences and high-performance coaching.

He has published more than 50 papers and patents at the intersection between physiology, health, technology and human performance.

He is the co-founder of HRV4Training and loves running.



Marco Altini

Founder HRV4Training.com, Data Science @ouraring Lecturer @VUamsterdam. PhD in Machine Learning, 2x MSc: Sport Science, Computer Science Engineering. Runner