Amaury Sports Organisation (A.S.O) the organisers behind Tour De France, along with Dimension Data, yesterday announced usage of machine learning in this year’s Tour De France to provide a level of insights and information unprecedented in this competition. But before we get into that, here’s some info about the Tour for all you cycling noobs out there.
Tour de France, for those who are not too familiar with the competition, is the oldest of the three most respected Grand Tours in the world of cycling. It was started in 1903 and has since evolved into a huge cycling spectacle that takes place over three weeks across France, with occasional forays into other countries. This year, the race comprises of 198 riders split into 22 teams, who will compete in the 3,450-kilometre long race in 21 stages spread across four countries – Germany, Belgium, Luxembourg and of course, France. The race covers diverse topographies and climatic conditions and poses a significant challenge to traditional telecast and coverage mediums like television. But all that has changed with the infusion of technological innovation into the sport.
Overview of the tech
With the help of GPS sensors placed in the bikes, Dimension Data is able to generate and provide live rider-tracking information, along with several other insights that are not possible through the limited visual tracking that on-ground (static) and aerial video feeds can provide. For instance, with location information for each rider, it is easy to track their speed, their formations, breakaway groups, moves and strategies – all in real time.
The GPS trackers placed in each bike relay the information to mobile relay stations as well the helicopters following overhead. This information is then relayed to ground stations placed near finish lines and checkpoints in each stage, where it is processed and further relayed to the cloud-based data centres where it is accessed by all the four teams working on the transmission and insight generation. Alongside the data analysis teams, the data is also available to television broadcasts and commentators, which enables generation of live-data overlays on the video feed based on the insights.
The data in question here is itself quite diverse. Essentially, there are six types of data that this system needs to process to generate everything that is required for the Racecentre to provide the insights that it does. They are:
- GPS/Telemetry data: Comes from trackers on the cycles in the form of a race situation JSON file. It contains the timestamp, latitude, longitude and the speed of the rider.
- Race data: Provided at beginning of race, Contains course details (sprints, climbs) as well as rider data (name, team name etc)
- Timing data: Data from official timing provider regarding results, photo finish etc.
- Environmental data: From third party sources, uses ordnance data as well hyper-localised weather data.
- Social Media data: from handles like @letour and @letourdata
- Media: Photos, videos and other data
There are two stages to the data transmission from the sensors. The GPS sensors in each bike (not the rider), has to relay the information to mobile relay points. The primary network, in this case, is a wide wireless area network (WWAN) based on 3G (3GPP 802.15.4). The accuracy is increased greatly by incorporating a mesh network between the GPS telemetry devices and the relay points, using the other GPS devices as reference points in the moving group.
The second stage involves sending the same data to end points. To be specific, from secondary relay points on the course to primary relay points in aeroplanes/helicopters flying overhead and on to the end of the race, where a ground station is established to process it further. This leg uses a licensed television RF broadcast frequency. After multiplexing the data onto the signal, the data is transmitted in a near-line-of-sight manner. To achieve that in the diverse terrain that the race covers, the receiving station is placed on a mobile lift about 40 metres above the technical zone at the end of the race.
The 198 riders cumulatively will generate 150 million unique geospatial and environmental readings. Alongside, there will be 35 data points per rider per second, along with 379,800 gradient data points, 11,394 weather data points, leading to a combined 3 billion-plus data points during the Tour. For a better understanding of the level of improvement, consider the fact that last year there were only 128 million cumulative data points.
Another area where data is transforming the sport are insights based on historical data and environmental data. Taking the environmental factors into account, for instance, the wind, rainfall etc in a certain area, along with historical data at the same location or for the same rider, the data centre is used to generate accurate predictions for the ongoing segment of the Tour and the overall Tour as well. These insights will be available at the Racecentre for the Tour, maintained by Dimension Data.
Christian Prudhomme, Director of the Tour de France, A.S.O. said, “Today, our followers want to be immersed in the event. They’re more digitally engaged on social media than ever before and want a live and compelling second-screen experience during the Tour. Technology enables us to completely transform their experience of the race.”
Sports – the new insights game
Alongside GPS sensors, the sport of cycling in general, and Tour de France specifically as well, use a lot of other sensors and trackers – like heart rate sensors on the cyclists, pressure sensors on the pedals and more. While these sensors aren’t being used for insights in Tour de France, the world of cycling, in general, is no longer unknown to in-depth insights based on a combination of all these data points.
In fact, sports, in general, has benefited a lot from techs like sensors and drones. The recently concluded ICC Champions Trophy 2017 was heavily tech-infused with Intel-powered drones and bat sensors offering a whole new level of insights about the sport. Going forward, similar insights are expected from almost every sporting event out there with a public appetite for objective, insightful performance data growing consistently as compared to the subjective performance metrics available so far.