Data-Centric Approach to Marathon Training

Daniel Corley
3 min readDec 30, 2019

--

I signed up for the 2020 Chicago Marathon as a joke after watching some friends participate… the joke’s on me, I got in.

With just under 300 days until race-day, I’m starting to think about training. And seeing that I’m in a data science bootcamp at Flatiron, I thought I’d bring some real-world analysis on how to do so into the picture, instead of following one of those ‘couch to marathon in 5 weeks’ guides.

Strava, an online running and cycling tracking app has published results for the 2019 Chicago Marathon, along with week by week training regimen for those that participated — found here. The training weeks are broken down by the runner’s goal times in 30 minute windows.

The Strava ‘Chicago Marathon’ page

To figure out the goal time group I’d like to analyze, I first had to come up with a goal for myself — I decided that I’d like to be in the top half of all finishers from 2019. After loading up the results into a chart, it seems that I’ll be targeting a sub-4 hour marathon (9:09/mile).

Starting to look at the data of the competitors that used Strava to track their results, a few things become apparent very quickly - 1) not all runners tracked all of their workouts, and 2) not all runners were able to beat their goal times in the marathon. Using those, I was able to create a filtered set of users that tracked workouts every week for the 12-week period, and had achieved their goal times. This resulted in 4 runners and 100 runs to sift through.

To give myself a fair idea of what my training should look like, I decided the best approach would be to take averages between these 4 warriors on a weekly basis, and look at number of runs, mileage per run, and overall total mileage. In doing so, I was able to create an overview of my new plan, in chart form:

Total Distance to Train Running per Week
Number of Runs, and Distance per Run per Week

If I were to follow this plan, I would see 39 runs with a total mileage of 349.47 miles over the 12 week period running up to the marathon. These would be split into an average of 3 runs per week, at an average of 15.90 miles per run.

To Follow Up:

  • In order to validate this plan, I would like to compare this to other training programs.
  • Will I actually follow this plan? TBD

--

--