Wednesday, January 7, 2015

The Power of Big Data

Peaks Coaching Group Power of Big Data

One of my favorite magazines is Wired, which is all about technology. Almost every issue contains some sort of article on human performance, which makes sense when you think about it; after all, pretty much all technology in the world exists to improve human performance in one way or another. From your cell phone to a CAT scan machine to the voice-activated GPS in your car, technology helps us improve ourselves, our effectiveness, and our productivity.

Several years ago Wired published a story by Chris Anderson entitled, “The Power of Big Data” with the subtitle “The End of Science.” The introduction read, “The quest for knowledge used to begin with grand theories. Now it begins with massive amounts of data. Welcome to the petabyte age.” Anderson welcomed us into the age where a massive amount of data thrown at any problem will eventually lead to meaningful and significant answers, the age where computers crunch more numbers than ever and algorithms applied to any large enough data set give us the correlations and patterns that science cannot. It’s this sheer quantity of data and the ability to process it that Anderson argued is the reason for the end of the traditional scientific method of hypothesize, model, and test.

This article really hit home for me, since the amount of power meter data we collect grows with each ride, each month of rides, and each year of rides. The power of big data really makes a difference for a power meter user. When taken alone as a few data points here and there, your watts or cadence data cannot provide you with any new insight. When you start to capture data every second of every ride, ranges of time become more and more interesting. A whole record of your workout starts to become recognizable as you point out the hills you went over, the attacks you made, or the intervals you did. You can compare sections within a workout to help determine whether you should have done five or six intervals or held a higher cadence, etc.

This ability is very enlightening. I’ve mentioned many times that I believe a power meter satisfies two of our primal desires as humans: First, it gives us a way of looking at our workout history and reliving our experiences. We all like to relive the “good old days” and talk about the time Bob attacked up that hill and then the yappy dog jumped out from behind the bush, bit Bob’s tire, and got stuck in the wheel going flop-flop-flop like Wile E. Coyote on Saturday morning cartoons. This reliving of experiences is made even more permanent by having a power meter file of your ride; it’s like having a second-by-second diary of your entire workout or race.

The second desire a power meter helps satisfy is the desire to learn from our experiences. With the power of big data at our fingertips, learning from our rides becomes more and more viable. The more data we have, the better. Even just looking at one individual workout and comparing intervals to each another allows us to learn from them.

The power of big data becomes even more valuable when we put it together over a longer period of time, or when we collect more data, making the data set even larger. An individual ride is great, but a week of rides is better, a month even better than that, a year even more so. Multiple years of data are amazing. I have one client who has used his SRM since 2003 and religiously recorded every single ride he’s done, which makes his data set quite large and highly valuable for analysis purposes. As his coach, I’ve been able to track and direct his progress over the years, learning about his specific responses to his training dose, pinpointing the type of training that brings him into form, and knowing with a high degree of certainty that his current training is correct. This sort of knowledge is priceless, and it’s not something coaches or athletes had been able to ascertain before power meters. Only with big data have we been able to make sense of the hundreds of rides athletes have done over time.

Let’s look at an example. In the screenshot below you can see an athlete’s mean maximal power curve over several years, which is one of my favorite ways to view changes in physiology over a long time period. There are two lines; the solid black line represents the athlete’s best numbers from all his rides in his current year, and the dashed line represents his best numbers from all his rides in the previous year. When TrainingPeaks WKO+ software goes through every workout for the time period, compares it to every other workout, and looks for the very best wattage values for each time period, the power of big data really starts to shine. The shape of the curve in the mean maximal power curve tells us about the athlete’s strengths and weaknesses, and by comparing curves from two time periods we learn how each physiological area has changed.


If you’ll look closely, you’ll see that the solid black line (current year) is above the dashed line in many places, indicating that the wattages were higher during those time periods and proving substantial improvement. When the more recent line is higher than the second line only in certain places, we can see points of specific physiological areas that have improved. Maybe your power at VO2Max improved or your sprint improved or your functional threshold improved. By analyzing lots and lots of data, you can see these trends and review the truth about what is really happening with your fitness.

Let’s look at another example of the power of big data. In this example we’re looking at the mean maximal power periodic chart, which shows your peak watts for specific times that you select. These are then charted over whatever time period you want to see, such as your current and previous seasons. In the chart below we see data for this athlete over a two-year period; by looking at the chart, it’s immediately obvious how much more consistent his wattage numbers are in the current year and how clearly he progressed throughout the season. Especially notice the steady progression of the green line and blue line (his one-minute and five-second bests), which relate to the anaerobic capacity and neuromuscular power systems, respectively. When all of the best wattages are charted for each week over a year or two of data, it’s much easier to see changes to the training regime, incremental improvements, and new personal bests.


The chart below is from the same athlete, but it includes only his data from one month, which is a smaller slice of the data set. We see that it looks like the one-minute (green) line and five-minute (red) lines have decreased, while the five-second (blue) line and sixty-minute (black) lines have increased.


Without a larger view and data set, true increases and decreases in wattage are hard to see. Since fitness changes over a longer period of time than just four weeks, it’s hard to say what’s going on by looking at only these numbers. Referring back to the year comparison allows you to see the bigger picture, the forest above the trees. That’s the power of big data.

When we combine the incredible number-crunching power of personal computers with the ability to record incredible amounts of data, the power of big data really does start make sense for us lowly cyclists. We aren’t curing cancer, calculating satellite orbital trajectories, or creating world peace, but cycling gives us that sense of personal satisfaction that is hard to find in other places in life. Using a power meter to help remember your experiences and learn from them can be deeply satisfying, while the power of big data can help you become even more successful.


Hunter Allen is a USA Cycling Level 1 coach and former professional cyclist. He is the coauthor of Training and Racing with a Power Meter and Cutting-Edge Cycling, co-developer of TrainingPeaks’ WKO software, and CEO and founder of Peaks Coaching Group. He and his coaches create custom training plans for all levels of athletes. Hunter can be contacted directly through PeaksCoachingGroup.com.

Originally published in Road Magazine
Image Credit: Shutterstock.com

No comments:

Post a Comment