Why Endurance Training Can’t Be Reduced to a Formula

Recently I tested a prototype of a wearable device that is intended to help runners monitor and control the intensity of their runs. During my back-and-forth email communications with the product’s lead developer, he sent me a link to a study titled “Intensity- and Duration-Based Options to Regulate Endurance Training.” The abstract began as follows: “The regulation of endurance training is usually based on the prescription of exercise intensity. Exercise duration, another important variable of training load, is rarely prescribed by individual measures and mostly set from experience.” Questioning the validity of experience as a guide to training prescriptions, the authors, a pair of Austrian exercise physiologists, went on to try to establish a more scientific method for determining how long individual athletes should train at different intensities.

The product developer who sent me the link was very approving of the Austrians’ approach. In his message to me he wrote, “When [this method is] paired with the individual’s aerobic and anaerobic threshold intensities, then a fairly complete training prescription can be applied to manage overall load and desired training outcome.” I was a bit more skeptical, replying, “Very interesting paper. So much of effective coaching is based in implicit knowledge and intuition. I think we’re very far away from coming up with a formula or set of formulas that can effectively substitute for these things, but as a coach I’m excited to see this line of research progress and further elucidate exactly what it is that the best coaches are getting right in their training prescriptions.”

I never heard from the product developer again. I believe he decided I was an idiot. If so, the feeling is mutual. Just kidding. He’s a very smart guy, but he comes from a mechanical engineering background, and like a lot of engineers he lacks a proper appreciation for complexity, and as a consequence of this lack he underestimates the power of experience and overestimates the power of theory-based predictions as a tool for solving real-world optimization problems such as endurance fitness development.

I use “complexity” not in the colloquial sense but in the scientific sense. According to Wikipedia (I know, I know), “A complex adaptive system is a system in which a perfect understanding of the individual parts does not always convey a perfect understanding of the whole system’s behavior.” Commonly cited examples of complex adaptive systems (CAS’s) are the human brain, developing embryos, and market economies. What all of these phenomena have in common is that even though they are ordered by relatively simple rules, it is virtually impossible to accurately predict their future states even if these rules are fully known and understood.

The inherent unpredictability of CAS’s presents a challenge for those who, for whatever reason, wish to anticipate or control their behavior. Although it seems reasonable to do so by studying the parts and rules that define the system, in practice this just doesn’t work. What works much better is trial and error, a reality that is very hard for engineers and others with a reductionistic mindset to swallow. And what’s even more galling for these folks is that a system really doesn’t have to be terribly complex for reductionism to fail.

Consider aircraft wing design. If you knew nothing about complexity, you might assume that the best way to design the most aerodynamic wing possible would be to use current knowledge of fluid dynamics to predict the design that is most aerodynamic, then build it, test it, and congratulate yourself on being right. In fact, the most aerodynamic aircraft wings in existence today were designed by trial and error because fluid dynamics—as simple as this phenomenon may be compared to an ecosystem—is too complex for reductionism to work.

To be clear, an airplane wing moving through air does not itself constitute a CAS, but it can be turned into one for the purpose of optimizing wing design. This is done through computer modeling. Engineers create simple programs that generate different designs quasi-randomly, test these designs in simulation, retain design facets that work better and discard facets that don’t work as well, then use this learning to produce a second “generation” of wings, and so on, until the process evolves the most aerodynamic wing possible. While humans are behind this process, they don’t really control it, and the designs they end up with are different from anything they could have come up with via the predictive method.

What does any of this have to do with endurance training? Heck, it has everything to do with endurance training! The problem of optimizing an athlete’s training for endurance performance is very much like the problem of designing the most aerodynamic wing possible. The physiology of endurance performance is extremely complex; place this physiology in the context of a living human being with thoughts, feelings, and emotions and you have something even more complex; place this human being in the context of a life with variabilities in work burdens, family stress, health, weather, and so forth, and you have something far too complex to allow any formula to correctly predict the training that will optimize an individual athlete’s fitness for an upcoming race.

This is not to suggest that training program design is always necessarily a complete shot in the dark. What I am suggesting is that, because of all this complexity, effective training is much more a matter of heuristics (learning and adjusting as you go) than of making great predictions before the process even begins.

Returning to the aforementioned study, I actually like the basic idea that the authors proposed therein. I think their tool could be useful for getting each athlete started on the right foot in his or her training. But I would caution against making too much of this tool or any similar one. Using it would not stop all kinds of surprises from popping up as the training process unfolded, and therefore all kinds of adjustments would be still necessary—adjustments that the tool itself can’t help with.

Nor would the tool even be necessary for getting an athlete started on the right foot. This can be done just as effectively with much less scientific tools. As a coach, all I really need to know is the athlete’s best and/or most recent races times and some basic information about his or her training history. By combining this input with my experience, I can design a program that will yield fairly predictable results. But the real work of coaching begins when the surprises come and I am able to rely on my experience to make adjustments in response to both on accidents that may never recur and to things I learn about the individual athlete’s body and mind, whose influence is recurrent.

For example, I might start off giving a certain athlete two easy days after each weekly interval workout and one easy day after each weekly long run, only to discover that this particular athlete, unlike most, recovers more quickly from intervals than from long runs, in response to which I will adjust his or her schedule to better balance stress and rest for this individual. And here’s an even better example: Often I schedule particular workouts at particular times primarily for the sake of boosting confidence, and only secondarily for some physiological benefit. I’m convinced this practice is effective. Can a one-size-fits-all prescriptive formula based on general human physiology do this? I think not.

Perhaps there will come a day when a computer can coach endurance athletes more effectively than an experienced human coach. After all, computers are already better than humans at chess. But endurance training isn’t chess. With its capacity for implicit learning, the human mind is uniquely suited to the job of training endurance athletes. I understand why folks like the developer of the wearable device I mentioned at the beginning of this post scoff at intuition, which seems so squishy and subjective and non-rigorous to the engineer’s mind, but it is immensely powerful as a tool for real-world problem-solving.

I am reminded of these passages from my book, Run: The Mind-Body Method of Running by Feel:

Malcolm Gladwell’s Blink: The Power of Thinking without Thinking is essentially a book about intuition. In it, Gladwell mentions another book, called Sources of Power, by Gary Klein, which discusses how high-performing professionals in various fields rely on intuition to make good decisions. Gladwell tells a story he heard from Gary Klein about a firefighter who thought he had ESP because he often knew what was going to happen on the job before it happened. One night he and his men were battling a kitchen fire when he suddenly ordered everyone out of the house. He did not know why; he just did it. As soon as they had escaped the house, the floor they had been standing on collapsed. No wonder the firefighter thought he had ESP! But Gary Klein’s in-depth interview with the firefighter, in which he was asked to recall every last detail of the situation, revealed that the firefighter had subconsciously registered various cues that the source of the fire his company was trying to put out was not in the kitchen itself but in the basement beneath them. Through experience on the job he had learned the patterns of different types of fires. And on that night his unconscious seat of implicit learning was able to recognize the pattern of a basement fire and deliver to the firefighter’s consciousness an urgent, intuitive feeling that he and his men were in serious danger and must flee the home immediately.

This story gives us an idea about how we should make intuitive decisions to build confidence through training as runners. The fireman who saved himself and his men by acting on intuition was, of course, extensively trained in fighting fires and brought a system of firefighting techniques to bear in fighting each fire. Nevertheless, most of what he really knew about fighting fires was learned implicitly through experience on the job. This knowledge existed in his unconscious as a capacity to recognize certain patterns before his conscious faculties did, make predictions based on them, and signal these predictions to his consciousness in the form of gut feelings. Similarly, every runner must learn and apply the principles and methods of training that have evolved over many generations as best practices. There are specific ways of training that are generally more effective than others for all runners, just as there are more and less effective ways to fight fires. But each runner is unique, and every day in the life of a runner presents a novel challenge in the quest to improve. Only by learning through experience can the individual runner gain proficiency in customizing his application of the proven principles and methods of training and in making good predictions about how specific training decisions will affect his fitness development. And most of this learning is implicit, as it was with the firefighter in Gladwell’s book. The runner’s subconscious faculties are usually first to figure out what the runner should do next, and communicate their conclusions to consciousness as feelings and hunches.

This has turned into a really long post. Sorry about that. Anyway, I trust I’ve made my point. Enjoy the rest of your day!