Goal-Directed, Principle-Guided Experimentation – 80/20 Endurance
Robot Coach

Goal-Directed, Principle-Guided Experimentation

I’m working on a new project involving artificial intelligence and endurance training that I could tell you about, but I would have to kill you. Just kidding—it’s not that secret. In any case, the project has got me thinking about fundamental questions in endurance training. For example: What is training?

Don’t snicker. The answer is surprisingly nonobvious. If I were to ask ten coaches to define endurance training, I would probably get seven or eight different responses, and they would be telling. Ten coaches who have seven or eight different conceptions of what training is are likely to coach athletes in seven or eight (at least slightly) different ways. After much pondering (in truth, it came to me in the shower), I’ve settled on the following formulation: Training is goal-directed, principle-guided experimentation.

Goal direction is what distinguishes training from exercise. Most people who exercise have some kind of goal, but one can achieve the goal of, say, keeping one’s weight under control by running for 40 minutes at low intensity every other day year-round. Exercise, in other words, is a fixed routine, like dental hygiene, whereas training is an evolving process. Exercise becomes training when you set a goal to achieve peak performance in an upcoming race. Doing the same, easy to moderately challenging workout over and over will not suffice to deliver you from the Point A of your present fitness Level to the Point B of optimal race fitness. Unlike exercise, training aims toward a specific destination.

Principle guidance is a set of tools and rules that are deployed for the purpose of getting the athlete from Point A to Point B. As part of the project I’m working on, I’ve taken some time to create an exhaustive list of the tools and rules that I use (unconsciously, for the most part) to train the athletes I work with. There are surprisingly few of them. Here are some:

Start where you are: The initial training load must be equal to or slightly greater than the athlete’s recent training load.

Purpose-structured workouts: Endurance fitness has multiple components that (for the most part) must be developed individually by workouts of different types that are structured specifically to fulfill a given purpose.

The 80/20 rule: Except in the early base (90/10) and taper (70/30) periods of training, the athlete must spend about 80 percent of their weekly training time at low intensity and 20 percent at moderate to high intensity.

Step cycles: The training process should be broken into three-week step cycles, in which the Week 1 training load is slightly higher than that of any preceding week, the Week 2 training load is slightly higher than that of Week 1, and Week 3 is a recovery week, where the training load is 10-20 percent lower than in Week 1.

The hard/easy rule: The more challenging a workout is, the more time should be allowed before the next challenging workout.

The foregoing principles, plus a few others, are sufficient to generate a complete, customized training plan for a given athlete aiming toward a particular goal. But the plan won’t be perfect, because the athlete is sure to respond to it in unexpected ways and unexpected events are certain to occur. The athlete may experience a week of heavy fatigue and poor performance, or suffer an injury, or gain fitness faster than anticipated during a particular period, or encounter any of a number of other eventualities that require the plan to be adjusted in order to keep them on track toward their goal.

Indeed, such adjustments are so inevitable that it is arguably unnecessary to create a plan in the first place. Instead, the training process can be treated as an experiment in which the next step is always determined by the results of the last step, and by the goal, and the aforementioned principles. As a matter of fact, as I’ve mentioned in past posts, I gave up planning my own training in any detail long ago, and my competitive results have not suffered as a result. In fact, they’ve gotten better. And I take the same approach with the athletes I coach.

The defining error of inferior coaches, in my opinion, is putting too much faith in planning. Athletes, too, for that matter. Everybody wants to believe they can know ahead of time where they’re going to end up, but you can’t really control that. What you can ensure is that you make progress in the general direction of where you’d like to end up, and this is best done by conceiving of training as process of goal-directed, principle-guided experimentation.

Can artificial intelligence do this as well as, or better than, a human coach? Not yet. The AI experts I deem most credible tamp down expectations, suggesting that in this context it will never do more than help human coaches do their job better. In the meantime, anyway, I’m at least having fun trying to put myself out of a job.