Our aim with this R&D project is to showcase our process through a live event, and cycling climbs provide the perfect platform. Our work frequently takes us deep into different (often complex) topics, requiring meticulous research, data exploration and prototyping, before the end product emerges.
Past and present
As the first notable climb of the 2014 race (aside from the amazing “Côte de Jenkin Road” on Stage 2 in Sheffield), the summit finish to La Planche des Belles Filles on Stage 10 looked a good starting point for this project for two reasons. First was that the race finished here in 2012 on Stage 7, so we had some historical data to play with. Second, it looked to be the first stage that would shake up the general classification, although Wednesday’s cobbles blew the race apart early, and sadly Chris Froome has had to abandon. Hopefully though, with pre-race favourite Alberto Contador now in 19th place, 2’37” down on the yellow jersey of Vincenzo Nibali, we should see some attacks on the climb.
Initial prototyping has centred around what data is available, in what format and with what potential. The elevation data we’ve used comes from the European Environment Agency, which itself is based on NASA’s Shuttle Space Radar Topography and Global Digital Elevation Map. We’ve sliced them into given areas using the Geospatial Data Abstraction Library. We’ve then taken the route data from the official event website, and mapped that onto our 3D model (the stages from 2012 and 2014 share the same final 12km).
Training Peaks’ social media co-ordinator Bryan Alders, has granted us access to physiological data from 2012, including Janez Brajkovič, who finished 8th on stage 7. Its difficult to be completely confident of the accuracy of the elevation data – we know the finish is at 1035m, while Brajkovič’s data reaches 1108m (quite a discrepancy even if he continued up the road to the team bus) – interesting further reading here. This process is trial and error, and all part of prototyping: digging for data and pulling out the most interesting sources. We’ve synced the Brajkovič data to the climb visualisation at the bottom of this post, and can observe fluctuations in his cadence (RPM), heart rate (BPM) and speed (KM/hr) over the route.
One excellent source of cycle climb data is VeloViewer run by Ben Lowe, which provides interactive visualisations for thousands of routes based on Strava data. The VeloViewer elevation of La Planche des Belles Filles clearly shows the steep ramp just before the finish line, where Chris Froome attacked and won in 2012.
By calibrating the route backwards from the finish, in 100m sections, we’ve been able to manually add key events onto the map from the TV coverage. Whilst not as ideal as having an official live data stream, it does provide a really good editorial tool, and has been the most interesting application to date. It’s not perfect, since the race distance marker counts down kilometres to the finish for the leading rider on the road, not necessarily the riders of interest (who may be further back). But acceptable estimates can be made based on the time splits at each point (this is a prototype, after all).
This prototype is the basis of what we’ll be using to track the race live on Monday (a few final bugs aside), including an extended section back to include the penultimate climb of Col des Chevrères. The map uses WebGL, so will not work across all browsers or devices.