My first marathon may have ended in a forgettable flurry of fluorescent exhaustion, death approaching me after 3:31.23 in the late-Summer sun, but the dream started in an unusual state: on a flight between Melbourne and Chicago via Hong Kong where I picked up food poisoning and ended up losing 17 pounds in three days, but not before I read a nice little article in Runner’s World on my iPad Mini.
There was nothing profound about the article, but the thought of running 42.2km during my state of self loathing and sympathy eased my struggle slightly in between trips to the plane bathroom.
I dabbled in a basketball career through my younger years (semi-professionally, as my colleagues like to state) but was coming off five months of relocating to the US from Australia, a booze-soaked European summer honeymoon, moving within the US twice in five weeks, and then ending up in hospital with the aforementioned colon infection. So planning to run a marathon at that point was as foreign as watching our data scientists build an algorithm.
The idea stuck, however, and not even a UK trip the day after my hospital visit or, more profoundly, a Chicago winter, could deter me.
My first thought was to do some research, find a training plan online, and purchase a GPS watch. Perhaps some running shoes and those energy gels that appeared so handy. But then an obvious thought struck as I realised I literally had the best athlete tracking technology in the world at my disposal, with a room full of sports scientists next door. (I would say they are at my disposal, but that’s not where priorities lay at Catapult).
This took me down a stubborn and polarised path of refusing to look at any training plans or using a GPS watch at all. I would train for my first marathon, I declared to no one in particular, using only PlayerLoad as a guiding metric. Time? Distance? Speed? Irrelevant.
PlayerLoad as a metric is a revelation for sports science. Developed by Catapult over a decade ago at the Australian Institute of Sport (AIS), PlayerLoad captures all the raw data from the device’s accelerometers and magnetometers, filters it through a complex equation (below), and spits out one number that shows precisely how much work that athlete has done.
The empowering feature of this metric, used extensively by our 750+ global clients, is that it works in every environment (it doesn’t require GPS) and is irrelevant of the sport being played. Think of it like a pedometer counting your steps in your pursuit of the fabled daily 10,000, mixed with the most advanced sports science laboratory imaginable, and you have a scientifically-validated number to communicate athlete biometric volume.
Its obvious value aside, PlayerLoad is an arbitrary metric that means little without context. As seen in the above equation, the true figure is divided by 100 - but you’re still left with a very high figure. Some examples of ‘average’ PlayerLoads in various sports:
- Australian football: 1200
- Basketball: 800
- American football: 700
- Ice hockey: 600
- Rugby: 1000
The averages within each sport obviously vary dramatically between positions, which is why variations within PlayerLoad can take your performance analysis to new levels if you divide the figure by time or distance.
Because I would just be running and not be getting hit or changing direction or jumping, I would use normal 3D PlayerLoad (an accumulation of movement in the three bodily planes).
So in early 2016, I took off running. Again, this was in Chicago. In January. And my running route of choice was along Lake Michigan, where a large chunk of the coast line leading up to the popular North Beach was literally eight-feet piles of ice. So I kept the running casual early on.
Despite ignoring speed and distance, I still needed to build some sort of dynamic periodisation model to ensure I’d be able to go from running for 10 minutes on a treadmill in January, to making it 42.2km in a reasonable time in October.
I’ll skip the day-to-day details and share some interesting findings:
- PlayerLoad varies somewhat significantly depending on the running surface. Concrete obviously had the highest readings, sand was low (but showed worrisome asymmetry data when analysing running gait and force produced by each leg, so I avoided it to mitigate the risk of injury - and because the sand was basically frozen), and the convenient running track along Lake Michigan was lower than concrete but higher than grass.
- The Nike shoes I eventually ran the marathon in were the most effective in terms of reduced load, and thus increased efficiency. I won’t name the different types of running shoes I tried, but load scores did vary between pairs.
- I lived in a 30-storey apartment building for most of the year, so when I had less than 30 minutes but wanted to get a good run in, I would go up and down the stairwell a handful of times. This resulted in a lower PlayerLoad per second than running on concrete, but this is likely a result of less foot contact time and less impact on each step since the movement is essentially not complete. However, going downstairs shows significantly higher PlayerLoad.
- I lied when I said I didn’t look at speed and distance - the device measures them better than anything so I’d be a fool to ignore it in the data. I summarised each run in terms of its volume (total PlayerLoad and distance) and intensity (average and top speed, PlayerLoad per minute, and PlayerLoad per kilometre). The system isn’t currently designed for the athlete - our user is the strength coach, sports scientist or performance analyst. Therefore real-time analysis is not available ‘on the go’. I grew to prefer this - I could run free (no music, that’s for the mentally weak) without the bumpers of performance data, but still have world-class data waiting for me when I got home.
- Predictive analytics is tricky to encapsulate in any field, but even harder in elite sport when it comes to injury prevention - ie. it’s difficult to say you prevented an injury if the injury never happened. For some reason, the teams we work with don’t like the idea of injuring their players so we can build preventative algorithms around them, but fortunately we’ve built up enough data that this will be a reality in the future. I mention all this, in what is becoming the most dense set of bullet points, because a couple months into casual running, I started to feel pain in my left calf. I’d gone from nothing to 10km runs in -20 degree celcius weather pretty quickly and my chicken legs weren’t ready for it. The cool thing? Upon glancing at my data one day, one of our sports scientists asked if I was having an issue with my left calf or knee. After getting past the “holy sh*t, our technology is awesome” phase, he explained that I wasn’t pushing off with nearly as much force on my left side, but was landing heavier than on my right side. Despite the discomfort, I genuinely didn’t think I was favouring it while I ran, so this was a revelation for me.
- The previous point was dramatically increased in the few days following each stairs session I did. It took me a while to see the correlation, but once I gave up the stairs for an extended period, my calf improved significantly.
- Running is such a quixotic past time. Like reading, even if you’re doing it with someone else, you’re really doing it by yourself. To do the same repeated motion for extended periods of time both teaches you how to be alone and tests you equal parts physically and mentally. Or maybe I’m just a weird loner that had too much time to myself. One thing we can agree on is that this dot point was in no way interesting or informative.
- My range for PlayerLoad per minute (i.e. intensity) was between 15 and 25. This is consistent with the ‘average’ of 800 in a basketball game since a 25 would be the equivalent over 32 minutes, a typical number of game minutes for a player recording such a number.
In total, I ran 850km in 2016. My longest run leading up to the marathon was 36km, I ran 21km twice, 32km once, and the rest of my runs were between five miles (on the treadmill, hence the change of measurement) and 10 miles.
- I’m not sure what traditional training plans suggest in the way of number and volume of runs per week, but when I actually started properly training for the marathon I cut back to two runs per week - 16-32km on a Sunday and 5-12km on a Wednesday. I never ran more than 34 miles in a week, and never covered more than 106 miles in a month.
Then October 11th arrived and I questioned why I never used a training program, why I didn’t buy a GPS watch, why I chose the Chicago Marathon to be my first ever running race.
Everything went swimmingly, of course. The warm weather pushed back my desired finish time and I ended up completely changing my outfit the morning of the race, but the distance proved easier than I was expecting and our sports scientists had some interesting findings:
- I made 36 low-intensity changes of direction. With 45,000 fellow runners and being in one of the final corrals, I spent a lot of time zigging and weaving around the walkers and people casually crossing the road with their large pink signs with re-hashed puns like ‘Chafing the dream’, ‘You run better than the government’, and ‘If it was easy it would be called your mother’.
- Related to the above point (the changes of direction, not your mother), I actually covered an additional 850m over the length of the course.
- My total PlayerLoad was 3885 - almost the equivalent of playing five NBA games. This means my PlayerLoad per minute was 18.4, and my PlayerLoad per kilometre was 92.
Assuming no one is still reading, and based on my proven performance data analysis prowess, I’m now essentially a sports scientist and am happy to answer any questions you have at email@example.com.