Finding marathon intensity
a data-driven approach
The marathon is a tough distance, maybe the toughest, as it is one of the few events that can be both long and high-intensity. We can run long, and we can run hard, but doing both the way a marathon requires … the difficulty might be one of the reasons why it is such a fascinating event - at least for me!
Over the years, it has been particularly challenging to find a sustainable pace. I speculate this is at least partially due to my unusually high energy expenditure (see for example the data I collected during my PhD, shown here, which translates into very poor running efficiency). This is not something that has much impact for shorter distances, but the gap between half marathon pace and full marathon pace for me is very, very large (with respect to what would be the average of other runners).
Despite training relatively high volume (90-140 km/week) and doing routinely very long runs (30-60 km), failing to maintain this gap results in terrible cramps already after 26-28km, and the much-dreaded outcome shown in the picture below.
These issues make all sorts of coach wisdom or marathon predictors provide very poor guidance (and quite some frustration), in my case. As a result, training and racing for marathons has been a long, challenging process, in which using heart rate data has eventually played a key role (together with some trial and error).
Below are some considerations on the approach I used, which I hope can benefit others in a similar situation.
Heart rate data to the rescue
First and foremost, as I’ve noted elsewhere, I normally do workouts and races based on perceived effort, or feel, and then keep an eye on pace. I do not currently have much use for heart rate data when doing shorter intervals, threshold sessions, or racing anything up to a half marathon (possibly because I’ve done plenty, and eventually learned to manage them by feel).
However, for the reasons detailed above, this approach is ineffective for me in the marathon, as I need to be more conservative, possibly due to my energy expenditure, high sweat rate, and sodium loss, which might be the cause of, or associated with, the cramps and other issues I routinely experience (cramps are not something that happens to me once in a while in a race, but something I systematically experience every single time I do a long run with some intensity - I link this to being an outlier in (in)efficiency). While the research on cramps is not particularly helpful, it is clear that I get depleted quickly, and this impacts my performance for the long distance. Hence, somehow managing this depletion and inefficiency becomes key.
In order to try to determine in an objective and reproducible way what intensity I could sustain I have used heart rate. Heart rate remains the easiest to use and most practical tool to capture internal load during exercise, and as such, I started looking at my data to see if I could spot clear differences between races and long runs in which I experienced issues and events in which I did not.
To do so, I used Strava’s APIs, and grabbed my workouts over the past years. Then, I separated successful long runs and marathons, and not-so-successful ones, which I have manually annotated as “imploded”, a category that includes races where I cramped really badly already before 30 km, resulting in quite a miserable day, or similarly, long runs where I had the same issue.
When looking at the data, a clear picture emerged.
Looking at the data above with respect to the “outcome” shows that heart rate clearly separates outcomes. This means that a slightly higher heart rate always resulted in poor outcomes and cramps, and therefore, heart rate becomes a key parameter to manage for me. We are talking about small differences here, just a few beats per minute, and yet the shapes are very well separated.
From the graph above, we can derive that aiming for 85% of my maximal heart rate or ~159 bpm for the first half of the race seems a safe strategy. This is how I picked my marathon intensity back in August, regardless of the pace it would correspond to.
Then, I started training it.
Training for the task
Apart from finding the right intensity that I could sustain, I also (obviously!) needed to train my ability to sustain that intensity for the duration of a marathon. Running at a certain intensity gets us more efficient at that intensity, and similarly, the principle of specificity tells us that the closer we get to our target event, the more time we should be spending at our target event intensity. This is why marathon training goes from polarized to pyramidal in terms of intensity distribution (I have to thank my friend Dave Lipman here for the many great conversations on our runs, that made me better understand the importance of these aspects, and marathon training in general).
For this training block (more details here), given that I had my goal heart rate, I parked my ego and went out for long runs at marathon intensity, regardless of pace. This is very different from how I unsuccessfully trained in the past, setting goal paces and routinely failing to achieve them (cramping, interrupting workouts), something that should have been quite clear when looking at heart rate data (as shown in the previous plot). But again, this is always easier said than done, as the marathon is a complex event, especially when we are inexperienced at the distance. During this training block, I built gradually but rather quickly (I have been running for many years after all), eventually settling on what is now my favorite marathon workout: 10 km high-end zone 2 (just around the first lactate threshold), and then 25 km at race intensity (~159 bpm).
I’ve done 4 of these runs (plus some shorter ones), as during this type of training block I keep doing hard threshold runs too (as opposed to only doing marathon pace), for reasons I explain here. Unsurprisingly, the pace was very far from what I was racing on shorter distances in the same period (3’48”/km 10 km, 3’55”/km half marathon, but running 4’30”/km as race intensity during these long runs). This was partially due to the climate as I was spending time in Barcelona, which was rather warm and humid, and partially expected given my characteristics (running inefficiency). However, I just stuck with the plan and did not bother at all about my pace. If this was the best I could do at that moment, so be it.
Fast forward two months and I am back in Italy for a few talks with professional teams using HRV4Training and at the Italian Olympic Committee in Rome, which I then combined with a marathon in my hometown, Ravenna. The marathon was three weeks after a half marathon that I ran in 1h 22’ in Spain.
On one hand, I felt like I was gaining confidence and making clear progress. On the other hand, my long runs were more or less the same pace, far from exciting. Back in Italy, in a better climate, I went for my last hard, long run, 7 days before the race (this workout). This was a big step up in terms of the pace I could produce for the same heart rate, and changed things a bit.
Putting it to practice on race day
In the previous plot with my heart rate data you can see a bimodal distribution (there are two clear peaks, separated by a gap). Given this distribution and the fact that at this point I was too close to a 3-hour pace not to try, I hoped to have some room for a slightly higher (but well-controlled) effort and therefore decided to flirt with implosion and run at 86% of my maximal heart rate. Below you can see the actual marathon in light blue, placed perfectly between the two groups of prior efforts, unsuccessfully (in gray, too high heart rate), and successful (in dark blue, lower heart rate):
Heart rate here matters more than any other signal because this is the hardest my body can work for this distance (in different environments and for different fitness levels, with some caveats). External load (speed or ‘power’) cannot capture any of this, and while important, cannot guide my marathon training.
If I had gone out at a certain pace or power for my workouts back in Barcelona, with a different climate, and before developing further my ability to run at marathon intensity, I would have probably experienced much earlier cramps, compromised training, and further adaptations, and eventually performed worse (as I had done for many years).
On race day, I did cramp at about 39 km, which is as good as it gets for me (when not careful in managing heart rate, I would get them at 26–28 km), and I was really happy with a 3h 02', and a fifteen minutes PR.
I was navigating a thin line on this day, staying near a pace I knew I barely had, and on a humid and windy morning, but I am happy with the decision I made. I was quite surprised to see that my heart rate stayed very stable even in the second part of the race, something I had never experienced before. The average shows that I was indeed slightly off target, and I could have probably stayed a bit more conservative, but again, the 3 hours were too close:
This was the first time since I started running marathons (2017) in which I felt like I trained and executed the race according to plans, and heart rate guidance played a key role in the process.
Back to marathon training
Last week I started again a threshold and steady-state block (steady-state is what I call marathon pace). During the previous training cycle (which I cover here), I spent a fair amount of time in a “VO2max block”, as I had previously gotten really slow while preparing for an ultra and suffering minor annoyances that did not allow me to do much high-intensity training.
In this block, as I am starting from a better place in terms of speed, I am dedicating much more time to threshold and steady-state training (details here). I am hoping that more specificity can lead to slightly better efficiency, and reduce the likelihood of cramps on race day (the next marathon for me is March 19th).
The approach I am using is the exact same: doing high-intensity work with threshold sessions, then doing long runs with plenty of marathon intensity. Keeping heart rate below 160 bpm is clearly key for me, and therefore I try to combine this insight with my perceived effort, to slowly try to become better at managing the intensity by feel for this distance as well.
Below is the first session of this block, which was the week after a 10 km PR. I went out and paced myself using perceived effort (especially on the hills pace is not particularly meaningful), and then kept an eye on my heart rate, which was a bit lower than expected, also highlighting the recent step up in fitness. Things are heading in the right direction.
That’s a wrap
Finding the right, sustainable intensity for a marathon can be challenging. On top of the typical difficulties of preparing for such an event, our individual characteristics will play a big role when we try to extrapolate from shorter distances. For example, high inefficiency changes the picture quite a bit for me.
Using heart rate data and looking at outcomes both during training and races was particularly helpful to determine what was a sustainable effort for me.
This does not mean that there are easy formulas or zones or percentages of heart rate that we can apply to determine what is sustainable for us, but likely we need some trial and error.
Over time, by combining the data with our experience and feel, we can train and perform better, or at least, that’s what I’m trying to do.
Thank you for reading!
Marco holds a PhD cum laude in applied machine learning, a M.Sc. cum laude in computer science engineering, and a M.Sc. cum laude 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 co-founder of HRV4Training, advisor at Oura, guest lecturer at VU Amsterdam, and editor for IEEE Pervasive Computing Magazine. He loves running.
Twitter: @altini_marco
I read this article, and about 5 of your linked articles. I've been flirting with the idea of doing exactly what you're talking about here for my first marathon this year. I wanted to run one last year but started getting some bad shin splints and such. I still got a half in last year before it got bad, but I've determined that I was definitely over training. Everyone always talks about pace because it's an easy and exciting number. Few talk about heart rate, and then even fewer talk about aerobic/lactate thresholds. I've been putting in some time reading on the subject and I LOVE the way you brought it all together here! It builds my confidence in my current approach, and I DEFINITELY related to the comment you made about your training being at the same general pace of 'unexciting'. I always want to push harder to make it more exciting, but I think this approach will work out much better in the long-term. Thanks for sharing your hard earned research and conclusions! It was an awesome read!
Have you any thoughts as to why your running might be inefficient?