#Trainerroad polarized training professional

Because machine learning methods are designed for the purposes of prediction, you cannot use them to “discover” the causes of Y, no matter how hard you try. The only thing you’re interested in is whether X carries some information about Y, not why it carries that information. If you want to use X to predict Y, it doesn’t much matter whether X is a cause of Y or whether the two are merely correlated. Machine Learning Doesn’t Answer Why-Questions Standard machine learning methods cannot answer these types of questions, although it may initially look like as if they could. In fact, it needs conduct this analysis for each possible interval structure. For example, to know whether 4x5min intervals are optimal for me today, the AI needs to evaluate what my performance would look like if I did 6x3min intervals instead. Moreover, to find the optimal workout, we need to evaluate what that individual’s performance would be under different alternatives. What we really want to understand is why that probability is 0.3. However, to be able to devise truly individualized training plans, it is not sufficient to know that someone’s probability of failing the next workout is 0.3. If predicting the probability of failing a workout was the only thing we cared about, then standard machine learning would be the right tool for the task. Presumably, if the probability is too high, you will be prescribed another set of intervals. For example, one of the ways in which adaptive training uses historical training data is to predict how likely you are to fail the next workout. Without going too much into the details, the modeling approaches discussed in the podcast focus on the standard goals of machine learning: classification and prediction. Here, I started wondering whether the machine learning methods discussed in the podcast are up for the task. The eventual goal is to tailor entire training plans based on users’ individual characteristics. As simple as this sounds, it’s already an improvement compared to standard stock plans where the next workout will remain the same irrespective of whether you can complete today’s intervals or not.įrom the lengthy podcast in which TrainerRoad announced their plans to roll out adaptive training (also discussed here), it’s clear they view this as an initial step towards truly AI-based training. For example, if you fail to complete a given workout, the system may change the intensity of your following workout downwards. They are currently rolling out what they call “adaptive training”, which leverages users’ training history and other individual information to adjust their workouts. No wonder, then, that companies such as TrainerRoad are looking to change things. Needless to say, the “one size fits all” approach discussed above is unlikely to produce optimal results. Usually, the only adjustment is based on a rough approximation of your functional threshold power. Moreover, each individual training plan is the same for all of the athletes that are doing it. If you are not working with an individual coach but still want to follow a structured training plan, your options are limited to stock programs that are all variations of a handful of basic approaches. Yet the methods for endurance training have remained the same. This means that there’s now an unprecedented amount of data on endurance training. What’s new is the ability of amateur athletes to do the same.

#Trainerroad polarized training professional

Of course, professional athletes have captured all of the above and more for years. When you do workouts, you additionally capture power, cadence, pedaling dynamics… the list goes on. You may also record various types of body measurements, heart rate data, body temperature, and so on. For example, if you are a cyclist who uses some kind of structured training plan, you typically capture some data of your daily activities, including sleep, movement and in some cases nutrition. In recent years, there’s been a sharp increase in the amount of data that amateur athletes can capture. The Emergence of Big Data in Endurance TrainingĮndurance training is ripe for disruption. No Amount of Machine Learning Will Tell You Why a Training Plan Works