4 Ways to Use the Training Data from Wearable Tech

For exceptional entry to all of our physical fitness, gear, adventure, and vacation stories, additionally discount rates on outings, gatherings, and gear, indication up for Exterior+ today
and save fifty %.

“],”renderIntial”:genuine,”wordCount”:350″>

The central concern that athletics scientists are grappling with these times is this: What the heck are we likely to do with all this facts? In stamina athletics, we’ve progressed from coronary heart fee displays and GPS watches to refined biomechanical examination, interior oxygen concentrations, and continual glucose measurements, all exhibited on your wrist then quickly downloaded to your pc. Workforce athletics have undergone a identical tech revolution. The resulting facts is interesting and ample, but is it really beneficial?

A new paper in the Global Journal of Sports activities Physiology and General performance tackles this concern and provides an fascinating framework for pondering about it, derived from the enterprise analytics literature. The paper comes from Kobe Houtmeyers and Arne Jaspers of KU Leuven in Belgium, together with Pedro Figueiredo of the Portuguese Soccer Federation’s Portugal Soccer University.

Here’s their 4-stage framework for facts analytics, presented in order of the two escalating complexity and escalating value to the athlete or coach:

  • Descriptive: What occurred?
  • Diagnostic: Why did it take place?
  • Predictive: What will take place?
  • Prescriptive: How do we make it take place?

Just about every stage builds on the earlier a single, which means that the descriptive layer is the foundation for anything else. Is the facts great more than enough? I’m pretty confident that a present day GPS enjoy can accurately explain how much and how rapid I have run in coaching, which permits me to transfer to the subsequent stage and attempt to diagnose no matter if a great or negative race resulted from coaching as well a lot, as well very little, as well tricky, as well simple, and so on. In contrast, the coronary heart fee facts I get from wrist sensors on athletics watches is utter garbage (as confirmed by comparing it to facts from upper body straps). It took me a though to know that, and any insights I drew from that flawed facts would certainly have been meaningless and potentially harming to my coaching.

Building predictions is more difficult (specially, as the stating goes, about the long term). Experts in a selection of athletics have experimented with to use device finding out to comb as a result of significant sets of coaching facts to predict who’s at substantial chance of finding hurt. For case in point, a review published earlier this yr by researchers at the University of Groningen in the Netherlands plugged seven several years of coaching and injuries facts from 74 aggressive runners into an algorithm that parsed chance based mostly on both the earlier seven times of jogging (with ten parameters for each individual working day, like the complete length in different coaching zones, perceived exertion, and duration of cross-coaching) or the earlier three months (with 22 parameters for every 7 days). The resulting model, like identical ones in other athletics, was significantly superior than a coin toss at predicting accidents, but not yet great more than enough to base coaching selections on.

Prescriptive analytics, the holy grail for athletics scientists, is even additional elusive. A very simple case in point that doesn’t call for any heavy computation is coronary heart-fee variability (HRV), a proxy evaluate of stress and restoration position that (as I discussed in a 2018 posting) has been proposed as a day by day tutorial for selecting no matter if to practice tricky or simple. Even nevertheless the physiology makes perception, I have been skeptical of delegating crucial coaching selections to an algorithm. That is a phony selection, nevertheless, in accordance to Houtmeyers and his colleagues. Prescriptive analytics delivers “decision aid systems”: the algorithm is not replacing the coach, but is delivering him or her with one more viewpoint that’s not weighed down by the inescapable cognitive biases that afflict human determination-creating.

Interestingly, Marco Altini, a single of the leaders in creating ways to HRV-guided coaching, posted a Twitter thread a several months ago in which he reflected on what has adjusted in the subject due to the fact my 2018 posting. Between the insights: the measuring technological know-how has improved, as has knowledge about how and when to use it to get the most dependable facts. That is important for descriptive utilization. But even great facts doesn’t assurance great prescriptive tips. In accordance to Altini, research of HRV-guided coaching (like this a single) have moved away from tweaking exercise plans based mostly on the vagaries of that morning’s reading through, relying as a substitute on for a longer time-time period trends like jogging seven-working day averages. Even with individuals caveats, I’d even now perspective HRV as a supply of determination aid somewhat than as a determination-maker.

A person of the good reasons Houtmeyers’s paper appealed to me is that I used a bunch of time pondering about these concerns all through my latest experiment with continual glucose monitoring. The 4-stage framework allows make clear my pondering. It is apparent that CGMs provide terrific descriptive facts and with some effort, I imagine you can also get some great diagnostic insights. But the gross sales pitch, as you’d expect, is explicitly concentrated on predictive and prescriptive claims: guiding you on what and when to eat in order to maximize performance and restoration. Possibly that’s possible, but I’m not yet certain.

In actuality, if there’s a single very simple information I acquire away from this paper, it’s that description and analysis are not the same factor as prediction and prescription. The latter doesn’t observe quickly from the previous. As the facts sets preserve finding larger and better-good quality, it appears to be inescapable that we’ll finally achieve the place when device-finding out algorithms can select up designs and interactions that even remarkably knowledgeable coaches might pass up. But that’s a significant leap, and facts on its own—even “big” data—won’t get us there.


For additional Sweat Science, join me on Twitter and Facebook, indication up for the electronic mail newsletter, and check out my e-book Endure: Brain, Physique, and the Curiously Elastic Restrictions of Human General performance.