Fast Talk, ep. 44: The data revolution — how A.I. and machine learning will make you faster

[related title=”More episodes of Fast Talk” align=”right” tag=”Fast-Talk-podcast”]

The VeloNews Fast Talk podcast is your source for the best advice and most interesting insight on what it takes to become a better cyclist. Listen in as VeloNews managing editor Chris Case and our resident physiologist and coach, Trevor Connor, discuss a range of topics, including training, physiology, technology, nutrition, and more.

This episode is all about data. Not long ago, people looked at you funny if you had a two-inch screen mounted to your handlebars. Now, we ride with head units the size of iPhones, sensors connected to our limbs, and wearables that track our every step and heartbeat. No one bats an eyelash.

A few episodes ago, we talked with Hunter Allen about the history of power and how we got to this point. Today we ask this question: Where is all of this data going, and what do we gain by covering our bodies in sensors like something out of a Star Trek Borg episode?

In this episode, we’ll first discuss the data revolution. There’s been exponential growth in the amount of information we’re now able to collect and analyze. That data is allowing us to analyze our training in ways we never could before, but it also comes with some dangers.

Next, we’ll discuss the rise of artificial intelligence and machine learning in training software. That’s a fancy way of saying that the software no longer just tells you how far you rode and what your average power was. Increasingly, it is telling you what your rides mean, where your form is at, and gives clues to what you should be doing.

Then we’ll address the three stages toward machine learning. First, there is the descriptive: how far you rode, how many hours you’ve trained this month, and so forth. The second, where we are now, is predictive: software crunches your data to predict what’s going to happen to your form. Finally, in a few years, we hope to move into the prescriptive stage where the software starts telling us what we should do.

Finally, since these changes are going to have a significant impact on athletes and coaches, we address what each can expect. Should coaches start refreshing their resumes?

Our guest is the lead engineer for TrainingPeaks and their coach-focused software package WKO4, Tim Cusick. He’s been coaching elite athletes for 18 years, including Amber Neben, a multiple-time national and world champion. He comes from the world of data analytics, which gives him a unique perspective on training science. He’s been working in A.I. and machine learning since the 1990s.

We’ll also hear from Armando Mastracci, the developer of Xert Training software, which grew out of Armando’s own experience as an engineer in crunching large amounts of data to find trends.

Dean Golich, a head coach at Carmichael Training Systems, will share his thoughts on where the software is headed.

And finally, we’ll touch base with Joe Dombrowski, of the EF Education First-Drapac WorldTour team, to get his take on how pros are reacting to the data revolution.

With that, let’s make you fast … and live long and prosper.

Fast Talk is available on all your favorite podcast services, including iTunes, Stitcher, Google Play, and Soundcloud. If you enjoy the podcast, please take a moment to rate and comment on iTunes after listening. Also, check out the VeloNews Cycling Podcast, our weekly discussion of the sport’s hottest topics, trends, and controversies.

References

Churchill, T. (2014). Modelling athletic training and performance: a hybrid artificial neural network ensemble approach. [PhD Thesis]. University of Canberra, 238.

Halson, S. L. (2014). Monitoring training load to understand fatigue in athletes. Sports Med, 44 Suppl 2, S139-147.

Jobson, S. A., Passfield, L., Atkinson, G., Barton, G., & Scarf, P. (2009). The analysis and utilization of cycling training data. Sports Med, 39(10), 833-844.

Kumyaito, N., Yupapin, P., & Tamee, K. (2018). Planning a sports training program using Adaptive Particle Swarm Optimization with emphasis on physiological constraints. BMC Res Notes, 11(1), 9.

Nimmerichter, A., Eston, R. G., Bachl, N., & Williams, C. (2011). Longitudinal monitoring of power output and heart rate profiles in elite cyclists. J Sports Sci, 29(8), 831-840.