DOE-ing myself: experimentation to better running
I ran an experiment on myself at the dawn of the pandemic. This is my story.
So this is me: I am a fairly healthy 36-year-old male living in Austin, Texas, USA. While somewhat active throughout my teens and twenties, post-marriage and kids has seen me become a bit too sedentary, a bit too lazy, and much too easily winded. In May 2019, my wife and I took our kids for a hike, with me bringing our youngest (at the time) in a backpack carrier. After 2+ miles of fairly level terrain, I was exhausted. This precipitated a sea change in my life.
On 27 May 2019, I started running. It was terrible. I hated it, yet I loved it. Three months later, I was talked into running a 10K trail race (at night), and I was hooked! Four months later, I heroically and quite stupidly ran a 50K trail race. It took me over an hour longer than I’d anticipated. I couldn’t walk for a week afterward, and when I crossed the finish line, it took every ounce of self-control to not openly weep.
Fast-forward to April 2020: I am ramping up for a series of four races longer than 60K over the next 10 months. Time to train is at a premium. So, as I am running, I am continually asking myself this question: What is the most efficient way to improve my running economy in a minimum amount of time? Enter design of experiments (DOE).
DOE is a mechanism to most efficiently use selected input variables to draw conclusions about pre-determined output variables. If it’s good enough to design top-of-the-line semi-conductor components, it ought to be good enough to help me run faster without running longer. So, what is my plan?
Well, there are some variables that I cannot control but I know are important: weather, mileage (I am using a prescribed training plan), hours of sleep, my weight, etc. These parameters are being tracked, just not controlled. But there are parameters I can control:
- Warm up — a 5-minute warmup on a rowing machine
- Caffeine — drinking coffee right before my run
- Timing of run — first thing in the morning or afternoon
- Shoes — trail or road shoes
- Diet — Did I eat meat in the previous 24 hours?
- Running hydration — water or electrolyte
- Apparel — 7”, 5”, or 2” inseam shorts*
*An ultramarathoning colleague and I have a pet theory that shorter shorts make you run faster. It’s time to test this theory once and for all.
I will be completing five runs a week for three weeks (see the table below). Through all my runs, I will try to maintain a heart rate in the upper end of Zone 2 to ensure the effort I put into each run is quantitatively consistent. For me, this means I will try to maintain a heart rate between 139 and 144 beats per minute (bpm), while trying to stay as close to 144 bpm as possible. I will also temporarily sacrifice my love of trail running and keep all my runs on roads/sidewalks to maintain consistency.
So how will I track progress? In addition to recording variables I cannot control, I will specifically track these three parameters:
- Grade adjusted pace (GAP) — pace in minutes per mile normalized for topography
- Average heart rate — while trying to maintain an average heart rate of 144 bpm, variability is sure to occur
- Perceived exertion — a subjective 1–10 measure of how I felt immediately after the run
An improvement in these three outputs is how I am gauging an improvement in efficiency — can I run faster, with a lower heart rate, and upon completing a run, do I feel like I exerted less energy.
Now, if I complete 15 runs over three weeks I will hopefully improve my running efficiency even without controlling these inputs, just by the sheer fact that I am running regularly. To account for this, my design uses week as a blocking variable. As a result, when I build the model upon completing this experiment, I will account for week-dependent impacts in running efficiency and subtract them out of the model. This will allow me to better estimate the impact of the factors am trying to study.
Results
Number of runs: 15
Total distance: 109.6 miles (176.4 km)
Total time: 15 hrs 6 min
Number of catcalls while wearing short shorts: 2 (5 if I count my wife)
Where week 2 was warm and humid, week 3 was cool and dry. While I began to feel the collective fatigue in my legs that comes with ramping up training, overall, I felt faster and more energetic (and the data bear this out).
The Take-Home Message
Timing matters! When modeling results based strictly on the controlled DOE factors, the most significant factors are ‘Running start time’ (earlier is better), followed by ‘Short inseam’ (shorter is better). If you learn nothing else from me today, it is that you should run early in the morning wearing short shorts. You can thank me later.
Consider the Effects of Uncontrolled Factors
In Central Texas, wild temperature and humidity swings between night and day are common. Throw in a random cold front, and the relationships become even more complex. Having grown up in San Diego, I can also tell you that the sun just feels more intense here. Could these factors be the reason that running start time matters?
Many more factors now have a P-value <0.05 but only meat consumption and caffeine are both significant and one of the controlled factors.
This is quite amazing. This tells me running is as much about the environmental factors. Timing, weather, and consumption (meat and coffee) exert the most leverage on the responses.
How well does this model predict responses? OK…
Uncertainty is substantial, and apparently if I become a vegetarian, I will regularly clock 7-minute miles. It seems to me more controlled runs are necessary, but this is a heckuva start and has pulled back the veil on a fairly nebulous process.
Uncertainty is substantial, and apparently if I become a vegetarian, I might just run faster (Scott Jurek would agree). It is clear the heat index plays a major roll (“Feels Like”) as the pace, perceived exertion, and heart rate are all elevated when it feels warmer. Also interesting is that a brief warmup makes the run reduces perceived exertion. It seems to me more controlled runs are necessary, but this is a heckuva start and has pulled back the veil on a fairly nebulous process.
Since completing this DOE I trained for four ultra marathons. I DNF’d the first, completed my first 50 miler for the second (Cactus Rose 50), had to back out of the third due to knee tendonitis (Bandera 50k), and ran a sub 24-hour 100-miler for the fourth (Rocky Raccoon 100). The DNF (did not finish) was a night race in Central Texas in late August. The starting temperature was near 95 degrees F with a heat index over 105 — after 27 miles, my body and will just gave up.
After this experience, and with what I learned through this DOE, I have set aside my love for morning runs and tried to embrace the Texas heat as much as possible, opting to run midday whenever possible. I recently made a second attempt of that same race (Cap’t Karl’s Reveille Peak Ranch 60k) and beat my goal time by 10 minutes and came in 5th place.
Sitting now, over one year since completing this DOE, the most valuable lesson I’ve walked away with is to embrace the discomfort, embrace the elements, run in any and all conditions you can, and trust that your training is enough — and of course, wear short shorts.
Portions of this post first appeared in the JMP Blog, where I wrote a four-part series on this experiment.
