Reading about someone else’s recovery-after-injury story can be a bit dull. At least that was my conclusion after pressing delete on my story a moment ago.
Having spared you the details, the summary is: I got injured. It hurt. It took me a year to recover because I didn’t tackle the recovery properly.
Measuring recovery
The point of this post is to think about how to measure recovery or how to assess running performance over a long time period?
One simple way is to take a metric such as pace and look at how it changes over time. If pace gets faster that’s good, and slower is bad. But there’s a problem with this. Slow paced runs may be slow for reasons unrelated to fitness. For example, a runner may focus for a few weeks on longer, slower paced runs to build a base. Without context, pace alone would suggest that performance during these weeks is poor.
So we need to take the course into account. Long or hilly runs will be generally slower than activities on short or flat courses. I described a way to compare pace on the same courses using a clustering method here. This method has the potential for looking at performance over the same course, but to assess running performance over a long period we need activities on the same course over that period, which we may not have. Ideally, we need to standardise running performance on these different courses and compare everything over time.
Standard pace
We can use “standard pace” from our similar course clusters and plot this versus time to understand changes in performance.
The plot shows the standard pace for all of the course clusters identified. Course clusters are those that are run five or more times, so we are missing data from courses run less than five times. The red line shows an averaged view of standard pace over time.
The break in point at the end of 2019 marks the point where I got injured. Well, more accurately, the point where accepted that I was injured and stopped running. I had been running injured through the latter part of 2019.
Recovery was slow. In Summer 2020 I felt like I had turned a corner and then standard pace has improved to (almost) pre-injury levels.
Technical details
The code used is under development here.
Standard pace is calculated as
\(z = \frac{x – \bar{x}}{S}\)Where x is pace for the activity, x bar is the average pace for the course (found by the clustering procedure) and S is the standard deviation of those activities. This means that performance is normalised to the course and is expressed in units of standard deviation (which is large for long, hilly courses, and small for short, flat courses).
—
The post title is taken from “You Better Run” by Epic Soundtracks from his Good Things LP.