How data analytics creates real-time race predictionsBlog
Here’s how we use data in new and exciting ways to revolutionise your viewing experience this year
For cycling fans following the Tour de France on social media and mobile screens around the world, we have special treat in store this year. We’re bringing you ever closer to the action, using advanced ways of combining and analysing data through the new algorithms we’ve built into our data analytics platform. These help us create real-time predictions as the race unfolds – a first in pro cycling.
Where does all our data come from?
We use GPS tracking devices installed underneath each bike’s saddle to track individual riders’ speed, position within the peloton, and distances between riders throughout each stage.
GPS tracking devices track the movement of all riders in the race
Third-party data give us environmental information, such as the gradient at which the rider is climbing or descending, as well as the weather conditions at that point along the route.
We also use historical data, which we’ve collated from our live tracking over the last two years, as well as rider performances, stage profiles, and race statistics across all UCI races over the past five years.
So what’s new this year?
For an even more revolutionary viewing experience, we’re adding machine learning and predictive analytics into the mix.
Machine learning is the true ‘brain’ of our computing capability. It’s the way in which our advanced data analytics platform correlates and integrates different sets and types of data. Through the complex algorithms we’ve programmed into the platform, the solution can then calculate the likelihood of possible race events in real-time, such as whether the peloton might catch up to breakaway riders or not, as well as create rider profiles based on their performances in past races.
Of course, no outcome is ever guaranteed … because anything can happen at a live sporting event. That’s what makes pro cycling so exciting to watch!
Here’s what you’ll get to see …
New rider profiles
We use historical data gathered from previous years’ races to create rider profiles. This helps us understand the particular environments and circumstances in which they perform best. For example, a rider historically inclined to do well across mountainous stages in the past, may be likely to perform well again during a climbing stage at the Tour, and so forth.
Data-generated rider profiles show a rider’s individual strengths and weaknesses against particular stage profiles
The #DDpredictor is the result of complex algorithms that analyse historical and live data to calculate the likelihood of real-time race events, such as whether the peloton will catch breakaway riders. You can also challenge the machine by making your own predictions and adding them to the social media conversation at #TDFpredict.
The #DDpredictor shows the likelihood of real-time race events during each stage
Live-tracking website and moment-in-time graphics
All of our live-tracking data sources feed into a real-time graphics display on the live-tracking website – the central online portal for all Tour de France race information.
These sources of data also help us tell a richer, more detailed story behind each day’s race action through moment-in-time visualisations such as heat maps, speed comparisons, and so forth – all of which you’ll see when you follow @letourdata. This year, we’re also introducing new television graphics which include 3D mapping, as well as replays overlaid with live data, such as the speed of the riders.
From comparative bar graphs to heat maps: you’ll see moment-in-time data visualisations on social media, television, and infographics with more information, more of the time