This article briefly looks at some great ways for us to take and use HRV readings to guide our training. Then I’ll outline some aspects of how I am using HRV in the lead up to my summer races.
Target Audience: Athletes not scientists
I first wrote about my own use of HRV in training 10 years ago and that coincided with my self-imagined glories in ETU/ITU Age Group Duathlons/Triathlons. Perhaps ‘leading-edge’ sports science at the time gave me a competitive edge?
I spent many years using HRV as one of the inputs to manually guide my training for the current day. Firstbeat Athlete and BioForce HRV were two of the most useful products that stick in my mind, the former was discontinued in 2016 and Bioforce HRV is still on sale, albeit in an improved format. Other apps later came into existence like the free Elite HRV and then there was an unusual product called HRV4 Training that claimed to take accurate HRV results using your smartphone camera. Perhaps foolishly I put the latter in the ‘nah‘ bin (I’m using it NOW and you’ll see why!)
Anecdote: I once totted up the amount of time I spent taking 3-minute waking HRV readings and was somewhat shocked that it added up to whole days…soon after that realisation I moved to an impressive product in 2016 called QS EMFIT that goes under your mattress and, as such, required ZERO time taking manual readings. It also had the massive advantage over a chest strap as it takes continuous HRV readings throughout the night. The downside was that I became mentally detached from the stats and, after a while, never looked at them. Sadly my EMFIT stopped working earlier this year.
Maybe 4 or 5 years ago the whole HRV ‘market’ took off as the understanding of HRV became more widespread. This was very significantly helped by wrist-based optical sensors finally being able to deliver accurate HRV readings at non-sporting levels of exertion. Resting-HRV is now routinely available in many sports and wellness products ranging from smart rings to the latest Apple Watch 6 or the latest Garmin ELEVATE sensor.
State-of-the-Art Consumer HRV Tech
Firstly let’s consider what measures HRV data can produce.
Consumer HRV tech can infer these: Stress/Readiness, sleep stages (with 65% accuracy), breathing rate, aerobic threshold.
But getting that data is fraught with difficulty.
The usefulness of HRV tech boils down to 1) the accuracy of the sensors, 2) the timing & method used to take a reading (context), and 3) the scientific correctness of the algorithms that are applied to the data.
The most accurate sensor is the Polar H10 using Bluetooth. In itself, the pod is great. However, the construction of the H10 strap has a large sensor pad area and nobbly dimples to stop the strap from moving – these two features reduce the chance of lost or incorrect readings. Similarly, Bluetooth will increase accuracy by losing fewer packets than ANT+. Don’t get what I’m saying here confused with accurate HR for your workout, the HR data from your ANT+ Garmin is accurate.
We will come on to the algorithms in a minute but quickly here is a chart showing the accuracy of some sensors plus their algorithms. You will see that the HRV4TRAINING app I mentioned earlier was the most accurate in the study and even when using a smartphone’s camera it remained actionably accurate.
The readings are all about context. I can’t stress that highly enough. A silly example would be where an optical sensor on the wrist gave a constant and correct rMSSD reading of 50 throughout the night…so you had a consistent night’s sleep? A: No!! Well, putting aside the impossibility of the series of readings, each reading represents different things in the context of each of your sleep stages. 50 in REM sleep is different to 50 in DEEP sleep and so on. Thus you have to understand the CONTEXT in which the reading was taken -sleep stage for example. When exactly did each of your sleep stages start and end? A: You will NEVER know for sure.
Then during the day, there is (are?) a multitude of factors that just make the context of your readings incomparable to other readings such as before meals, during exercise or after office work.
Context – Best practice
Each morning before you start your daily routine take a SINGLE READING, like this: go to the toilet, adopt the same position in bed as yesterday. Relax and take a 3-minute reading with a Polar H10 and your HRV app of choice. Then, and only then, get changed and go for your coffee and breakfast. Consistency is key.
Context – Not-so-bad practice
It’s the same as best practice but you could use a 1-minute reading and even that optical camera from HRV4 Training will give you a sufficiently accurate result, saving the purchase of a $70 chest strap.
Context – Other sensible methods
In parallel (but not instead of), use a smart ring or EMFIT to check the progress of your recovery throughout the night. Garmin Body Battery is probably OK for this in an indicative way too.
Context – Unsensible Analyses
The interpretation of continuous readings or periodic manual readings throughout the day is probably not based on science. Perhaps continuous readings might show a trend and perhaps they might show something interesting if contextual data is overlain like the results of a workout. I use a few apps that do this for ‘a bit of fun’ …that’s all.
The algorithms that process the raw data can be HIGHLY complex. In the most basic sense, they have to remove ‘noise’. You might simply think of this as working alongside the device’s accelerometer to determine when, for example, the wrist is highly active and perhaps the HRV readings at those time are discarded. However, more complex algorithms will use machine learning (ML) to determine how to treat your data based on patterns in ALL your previous data. ML is probably also applied from inferences drawn from population-level data.
Then we come to how apps apply the algorithms. Some work better on subjects with a lower rMSSD. Some choose random periods at night in which to take a representative nightly reading. Some take periodic readings usually hours apart. Some provide nightly average readings or use the lowest rHR/highest HRV. Some take sufficiently accurate manual readings but have almost useless automated readings between them. Some tech uses rMSSD calculations and some uses SDNN (each is fine but most of us use the former and one can’t be converted to the other without the raw data).
From a user perspective, it’s hard to know which technology is best. Some people point out that, for example, WHOOP is not a great HR sensor for broadcasting live HR data (which vendors’ is!) but, actually, some of its back-end algorithms are powerful and the cleaned-up resting HRV data is probably fairly good to the extent where Whoop’s SLEEP STAGE predictions are as good as the competition. Which doesn’t seem to make intuitive sense but, hey, #PeskyScience.
Putting HRV to Use
I don’t think HRV is generally as useful as many people now believe.
It seems to me that many products and metrics are being created that use HRV but which are built on shaky foundations.
That said, there are still some great uses founded on proper science.
I keep banging on about this.
Sleep stage analysis is a VERY popular feature for athletes, body-hackers and the tech-adorned population at large.
HOWEVER. The best accuracy of any consumer device in determining sleep stages compared to the gold-standard polysomnography is 85%, to score higher than that requires the agreement of 2 independent observers. Some of the best consumer tech can give 60-something percentage accuracy...that’s rubbish right? A: Yes.
ALL consumer-grade sleep stage estimation is rubbish in my opinion. Yet, like many of you, I still look at my sleep stages!
Subtleties to the signature of your heartbeat can be used to understand at which point you took a breath. You could then build on this to help understand your breathing rate, oxygen intake, VO2max by one method, and ventilatory threshold. There are also even cleverer deductions and tests
Aerobic Threshold Estimation via Indoor Ramp Test
I wrote recently how HRV can be used to determine your LT1/aerobic threshold in controlled, indoor conditions using a statistical method called DFA-a1.
Recovery – Are you ready to train?
Readiness to train has been one of the areas that some sports scientists have grappled with for years. My opinion is that the technology has significantly progressed and we do have some methods that can give us useful guidance – but there still is much more to be learnt.
Readiness to train is important to understand as the major sports training platforms all move towards ADAPTIVE TRAINING. ie where the training adapts to your personal situation. In the past coaches might prescribe one rest day every Monday or one rest day every 10 days or, for a veteran athlete, they might prescribe 2 rest days a week. All of those prescriptions could be wrong for you, for example, you might have excellent genes/physiology, you might be a lifetime athlete or you might have just slacked a bit over the last 6 days and have more to give tomorrow rather than needing a rest day.
Example of adaptive training
- Macro adaptation to your weekly schedule
- Micro adaptation within a workout as you fatigue (see Xert for some of the best examples)
- Dynamic adaptation to morning readiness readings from HRV
- Adaptation to personal outcomes with machine learning based on HRV. Correlate the OUTCOMES of your historic data set with your HRV readings before each workout then adapt your personal plan to choose the best workout for you based on today’s HRV (the AI endurance platform is moving towards this). Similar ML principles can be used to adapt your training based on the outcomes from large peer-group data sets.
But then; even if the readiness to train argument is settled, we still will not have measured actual adaptation by HRV and that might always remain in the realm of performing specific tests to demonstrate that you have improved.
Some inputs to my 2021 training
I still find a classic TSB model useful. This complex chart shows the load from the 3x triathlon sports (blue) generally all increasing since about Nov 2020. The -ve green line shows my lack of readiness to perform but that’s kinda good whilst training, I use TSB in one way as a warning – for example, it’s now at about -20 for me and that can be when I become prone to injury. In the last 5 days, I’ve completed two 8-hour rides. Common sense tells me it’s time to take it easy and so does my chart.
As you can see, all my red and blue lines above are trending upwards and that’s supposedly good. But really …is it? I just feel either ‘a bit’ or ‘a lot’ tired all the time and I’m trusting that my body is adapting. But is it? I’m not getting any younger and almost certainly I will not be recovering as well as I did 10 years ago.
So I have reverted back to taking consistent waking-HRV readings. I use HRV4TRAINING Pro and a Polar H10 every morning lying down in bed before I get up. The following chart shows my HRV over the last month. You can see that the shaded area represents my optimal HRV zone. that shaded area has trended upwards over the last month which is a good sign that I am adapting to my training. Also, importantly, I am mostly keeping my HRV in this range as shown by the dark blue line.
After I take my morning reading I also use parameters in the app to tag how I feel, such as for muscle soreness and motivation. I also tag factors like my perceived performance level the previous day and if I had alcohol. As your personal data set increases in size, correlations can be inferred between the contextual tags you have given to all your readings. Remember what I said near the start…ALL the clever stuff needs context.
HRV4Training then also drags in my HR data from STRAVA to determine my load – like on the first chart shown further above. I haven’t shown HRV4Training’s load chart as the maths will take some time to get the chart to look correct and the one further above is essentially correct.
Note: I have created a new strava account to which I pass complete and correct HR data. I link this to HRV4T and currently maintain the strava data manually. My other accounts have duplicates and that will mess up the training load data once the algorithms catch up.
Usefully, after each morning reading, I get the train/no train/take-it-easy recommendation. The following image shows that my 7.0 reading from this morning is within target and on the baseline and therefore I should train if I had planned to.
Normally I would have had a rest day today as I did an 8-hour ride yesterday.
Over the last 3 weeks, HRV4Training has validated two of my training choices
- Split a Tuesday swim+run so that the swim replaces a Monday rest day and the run stays on Tuesday.
- Split a Sunday morning brick session of hour bike +long run into separate morning and evening workouts. (I lose the serial multisport effect but get the load)
Today marks the start of my PEAK period. I’m pretty happy that I’ve handled the high load levels of 15-20 hours/week generally very well. I think the weeks to come will be tricky as the intensity now starts to get tough. I am anticipating that there will be times when I will take complete days off training guided solely by HRV4Training.
What’s New for me?
My long-term regular readers might say…yeah, but that’s exactly what you were doing 10 years ago! And, yes, it kinda was! However, the tech is better now on many levels. Here’s a brain dump of positive tech goodness on what’s improved
- The H10 sensor pairs first time every morning (Jeez, that used to be REALLY annoying and waste vast amounts of time). Maybe that’s iOS vs Android or an improved Polar?
- A 1-minute reading is sufficient, I used to do 3-minutes.
- The HRV4Training results/algorithms seem more correct than the ones I used to use.
- There are many more presentational and correlational features in HRV4Training which, importantly, can help better understand the context.
My use of HR & HRV with strength Training
Hmmm. It’s hard to interpret a combined HR load across swim, bike and run. Even if I have my correct LTHR/LT2-based HR zones for each individual sport can you really combine the loads from upper-body swimming and lower body running? Probably not is the answer (but I still do it).
I have a running power meter (Stryd), a swim power meter (Vasa) and several bike power meters, so I could theoretically combine the loads from the 3 sports. But the combined result would still be as dubious as one created from HR.
When we then add in a fourth discipline, weights/strength training, then I have the problem that my HR simply goes nowhere near the same levels as for running. I can sometimes get my HR into zone 3 whilst working my quads ….but not for long. Consequently, in my case, the HR load from weights is very very low. I doubt that those of you who can sustain their HRs for longer than me will have any significant load according to HR.
Yet strength training clearly does place a strain on the body. If you perform it less often than you should then the strain could be even greater after any given session.
Readiness-to-train from HRV can probably give more insight for strength training (read this). I would temper that by saying that I find RPE methods have value and that HRV tends to suggest I am ready to train before I think it is correct to do so (others say the same), although this might be because the strength training we perform is relatively ‘novel’ for our physiology. IE a rugby/American football player doing weights 5x a week would probably get more benefit from HRV guidance around weights than a runner doing weights 1-2x a week.
Where is HRV heading?
I’m not exactly sure where HRV is heading, it seems to be heading there more slowly than I expected! I can’t see too many Eureka moments coming, just incremental improvements and incremental adoption.
- Pro-training platforms (TR, TP, etc) – 10 years ago I had envisaged a much more widespread adoption of HRV by the training platforms by 2021. However now it seems clear that in 5 years time there will finally be high-quality, adaptive training available online from multiple sources linking to HRV readiness metrics.
- Consumer training guidance – I suspect that the majority of people never train sufficiently frequently or sufficiently hard enough to need adaptive training guidance from HRV. Furthermore, I can’t see many people in this category taking the HRV readings in the first place. However, we are seeing ring-based and watch-based technologies giving general advice to train or not to train and that simple decision-making piece of data will become more widespread.
- You’ve probably read of various HRV studies being commissioned to ‘predict’ the onset of Covid (#sigh). I think that is fanciful, however, my suspicion would be that the Covid example will be played out many times as consumer-grade devices and apps try to sell their wares to consumers based on spurious associations.
- We will also see the continuation of the trend for smaller apps that are focused on certain sporting and wellness niches. Some of these will no doubt be scientifically founded others will just look pretty and say ‘HRV’ several times in their marketing literature.
- Ultimately the link between HRV and ‘stress’ is generally accepted. I reckon we will get better at more accurately measuring that link but not that much better and determining what kind of stress is the cause.
This article is not sponsored in any way. I get nothing if you buy HRV4Training. I did get a free subscription from Marco
- If I’ve whetted your appetite for this, here is a more convincing post from HRV4T on why & how you should use HRV.
- Elite HRV – v2.0 Released
- Elite HRV – HRV Course
- Whoop’s Take on HRV
- Garmin-Firstbeat Metrics