Race Analysis: Lošinj Men's DH World Cup Race

As a data nerd myself (well my job title officially says analyst), I find meaning and tell stories with data and loved seeing this applied to my favourite participation and spectator sport. In the constant pursuit of evolving my skills, I have recently been brushing up on creating charts with Python. When learning I've found it helps if you're working with something in which you're interested so I looked to MTB and the World Cup, and remembering the old Geek Stats I set about testing myself to create faithful reproductions, with a few personal tweaks. After reaching out to Mark and getting his blessing – he's no longer writing the aforementioned stats – I am presenting my results here and hope to keep putting these write-ups out soon after each round of the DH World Cup, and maybe other race series too.

My work is all done in a couple of Jupyter Notebooks (though I'm working in the new Jupyter Labs), and is available online if anybody is interested in playing with this themselves. I've tried to clean up and optimize most of the code and it will be easily adaptable to other race datasets (at least from the same data sources) but some of it is a work in progress so excuse a certain amount of disarray. Technical details at the end.

---

On April 22 the 2018 UCI MTB DH World Cup series got underway for another exciting season, and this year we were kicking things off with a completely new track and venue in Lošinj, Croatia. Being a first time track we had no race history, which, combined with it being the first race of the year, meant we didn't have any real indicators of what to expect of the results, but we'd soon find out who'd been working hard through the long off-season.


Sector Analysis

All the pre-race talk about Lošinj was about the rocks. All the rocks. The first full-track preview video we were treated to also had a set of fast jumps at the top, and a short street section at the end. With the track split into five sectors, the first and last were basically the jumps and street respectively, leaving the middle three as the most potential for making or losing time, with sector 3 having the awkward triple-double combo. Despite the potential brutality of the rocky sectors, at the end of the day only one rider, Brook Macdonald, fell victim and was DNF. Due to the way UCI provides their data the sectors he did complete are not included in this analysis.


The Jumps



Macdonald in the Middle

Gwin had a roaring sector 2 that basically won him the race. He put a full second into the next fastest, Minnaar, and, given he was 0.2s back on them after split 1, put almost 1.5s into those to his eventual left and right on the podium, though all three were again in the top 6.


Sector 3 was the only of the day won by a non-podium finisher, Loris Vergier; something shared between the top 3 here. After having a clearly troublesome second sector, this brought Loris back into the top 20 for the third split.

Lucas and Gwin both held form through sector 3 rounding out the top 5, but Shaw's 11th place lost 0.7s to Gwin which is almost the exact amount by which he eventually lost out on at the finish, as we'll see later on. It was here between sectors 2 and 3 that Brook had his big off.



Double Bullseye

After winning sector 1 then struggling through the middle, Luca Shaw turned it right back up in sectors 4 and 5, where, in winning both, he clawed back 0.4s and 0.8s chunks of time against Golden Gwin, almost the same amount he lost out to him on sector 2. Without that bobble in sector 3, Shaw could have made it a very close finish. Gwin had his worst sector of the day at the end with an 11th giving up almost 0.8s to Shaw.

Sector 4:

Sector 5:

---

Heatmaps

---

Bonus Round

As I said earlier, I started this to play around with Python and Seaborn/Matplotlib so I decided to knock up a few other charts as practice and to see if there was anything interesting to be found.


The top placed guys aren't necessarily the fastest. Confirmed by this scatter plot of all 60 racers' speed against finish position.



But the two fastest guys are way faster than the rest. Yup, there's a definite correlation between start position and finish position. Weird huh?



Age (and I infer experience) pays in this game, and without the flat Minnaar would have tilted that line a little further. Britain and France have a raft of riders competing throughout the age range, which, given the chart we just saw, likely sets them up to continue strong results now and in the future.

---

Final Thoughts

So definitely some interesting tidbits in amongst all the data. There's still plenty more we could analyse but that's all I've had time to do for now, but they are fortunately all quickly replicable for future rounds so expect them in a timely fashion.

What else would people like to see, using the data available? Let me know in the comments, and I'll see if I can add it for the next rounds, starting with Fort Bill.

I'd love to be able to run wheel size vs speed and wheel size vs position, and especially compare those race-on-race to see how they suit perhaps some tracks over others. I do not have that data available however so without careful scouring of all the articles and race video, I don't think it'd be possible.

I will also try to get a Ladies write-up completed soon as their racing deserves the same scrutiny.

Additional analysis will be written up on my medium once this blog post is locked.

---

The Science Bit

Everything is public on GitHub, updated this evening with the latest notebooks, within which are notes on the process and instructions to run it yourself.

https://github.com/domwrap/mtbgeekstats

Data Sources:
http://prod.chronorace.be/api/results/uci/dh/race/20180421_dh/3
https://www.rootsandrain.com/race5897/2018-apr-22-mercedes-benz-uci-world-cup-1-losinj/results/filters/m/

---

Author: Dominic Wrapson

I discovered mountain biking and trail riding in summer 2008 on England's North Downs and by the end of that year had broken a leg riding DH. Just 7 months later in June 2009 spent three months riding and getting qualified to coach in the Whistler Bike Park, returning 9 months later to live "for a year" and never left. After many great seasons coaching and riding Whistler's incredible network of trails and trying to tick them all off, I recently became a North Vancouver resident with a goal of ticking off all those too.

You can follow and find Dominic on his Instagram, Twitter and plenty of other places by searching @domwrap

Special Thanks
Mark Shilton for the inspiration http://lookatthestats.blogspot.ca

---

Editor's Note: Now, with all that information, we're guessing you may want to take a look at your Fantasy Downhill Presented by Trek team again!