Tour de France riders’ saddles are for? They’re not just timing chips. They’re the sharp end of a partnership between the ASO and NTT – collecting over 150,000 points of in-race data per hour that’s transferred and transformed into the most advanced & enhanced race viewing experience we’ve ever seen – on television, mobile & desktop screens so fans can view anywhere there’s an internet connection.
This year we saw more cool graphics, more rider and race data, speeds, power output, and even predictions on when the break will be caught and who will win. That’s all thanks to NTT (formerly Dimension Data) – a top global technology services provides who specializes in you guessed it – collecting, crunching and applying data.
Click the video to watch the interview on the PEZ Youtube Channel, or scroll down to read the transcript.
But how does it all work, and what further race viewing enhancements can we look forward to next year? I talked with NTT’s Peter Gray – Vice President of Global Advanced Technology for Sport – himself a rider, Tour fan, and huge data geek, about how capturing over 150,000 points of data per hour is transforming our Tour viewing experience – and other parts of life as well.
Here is the full video transcript:
All right, Pez fans. Thank you for joining us once again. Today with me live on Pez via Skype from Melbourne, Australia, is Peter Gray. And, I’ve got to look at my notes for this, because his title is pretty big. He is the Vice President of Global Advanced Technology for Sport, for NTT. And, if you’re not familiar with this company, and if you’ve been watching the Tour, you will have seen some of the deepest dive into data we have ever seen from cycling. That data is provided by NTT, it’s sliced and diced and delivered to us in bite-sized chunks that we can actually use, and we’re going to get into that right now with Peter. Peter, thank you for joining us. Pleasure to have you on the show.
What NTT Does At Le Tour
I have been following the tour myself since 1986 and watching it religiously every summer. As I looked into the website and all the data that NTT are using to enhance the fan experience, it was quite incredible. So, Peter, can you give us a little bit of an overview of what NTT does and what you’ve done at the Tour this year?
One of the big challenges with road cycling is that for most of the population, aside from those of us who are really keen cycling fans, for most of the population, the Tour de France is the only bike race that they ever watch. And so, the challenge for the ASO is how to help make that understandable for people. And so, a lot of the work we’re doing is around how do we take all of that data, and there’s lots of it, but turn it into something that is useful and understandable for people at home? So, as we put together all of the television graphics, a lot of it’s about how do we best tell those stories, and then, which pieces of data do we choose to use and which do we choose not to.
We actually began working on this five years ago. So, we started as Dimension Data in 2015, began by starting with capturing information from the riders. So, putting sensors onto the bikes, getting GPS positions and speeds every second, being able to put that information into the real time broadcast, into the live broadcast. So it means on the TV broadcast we can get things like live time gaps, not just for the main groups, but for all the groups, and get time gaps to individual riders. You can get all of the speed information, all that kind of stuff as well.
So the evolution, really, has been from 2015, basically, putting these devices onto the back of the bike and getting data back from each of the riders, through to today, where we’ve got full integration of that data into the live broadcast. We’ve got machine learning and AI built around making predictions around the race. We’ve got the digital platforms, so you’ve got Race Center particularly, which you can go to and get live information about the groups, the gaps, but also journalist information. Twitter, video, audio, et cetera, and you can go and follow individual riders on Race Center and see exactly where particular individuals, or where all the riders from a particular team or from a particular country, you can do all that kind of stuff. And then we also created the @letourdata Twitter platform, which allows us to do in-race live data commentary using all that information. So basically, over that five years, we’ve continued to build that out year on year. This year we do it as NTT because Dimension Data has now become part of this big NTT organization.
How is data Collected & Sent?
You’ve got in your hand, I saw that electronic device, which we have seen show up on riders’ bikes now for the last few years. Tell us just a little bit about how that thing works. Where is the data being sent to and how is it being collected?
Basically the device attaches to the saddle rails on the seat, and within here you’ve basically got a GPS chip, transmitter and a battery. This is transmitting a position and speed every second, it bounces that information from the sensor to gateways that are on the television, motorbikes, on the officials’ cars and also on the helicopters and the airplane that follows the race. So, there’s an airplane that actually circles above the race, and it basically collects all of the television signals, race radio and the data from these, and bounces that information back to the race finish line. The zone technique, which is where all the broadcast is put together. And so, at the zone technique, that information comes back to us, and that allows us then to do all the data processing and analytics that then feeds that information into the TV graphics systems, into the websites, and into our analytics databases that then allow us to do that sort of live race analysis.
How is NTT set up to collect all that data and then, how many people are involved in this at the finish line, in crunching this data and getting it out?
We’ve got basically a control center at the finish line, which is a big double-story truck. Within there, we’ve got a small team of us who are coordinating all of the activities, both onsite, but also, importantly, off site. So, we have a team of about 30 people around the world involved in supporting the whole solution, but only a small team of about four of us on the ground.
On the ground, we’re really connecting with all of the different parties that we need to work with. The television team from France TV, the RF transmission team, so, they look after all the transmission of the TV signals back to the zone technique. Swiss Timing, they’re the official timekeeper for the race, and so we take all of the Swiss Timing data and results, feed that into the digital platforms, and the ASO themselves, who are the race operators. And so, their technology team base themselves out of our truck, our control center. And so basically, from that vehicle, we’re connected up with all of those different parties on the ground. And then we have teams working remotely in Johannesburg, in Melbourne, in the US, in India, in London, Paris, all working as a single team, delivering different parts of the whole solution.
Are those teams all working in live time? So, it’s race time in France, it’s middle of the night on the other side of the globe.
It’s the Melbourne team who really cop the worst of that, because, the race starts late evening and goes into the early morning. Not so bad for the Johannesburg team, they’re in exactly the same time zone as Paris. But, no, we’ve got a team who are very dedicated and passionate.
Predicting The Future With AI & Machine Learning
I’m curious to know more about the ability of the learning from the data you’re collecting to make race predictions. I’m reading some of the notes here on our interview talking about predicting key potential moments in the race, predicting stage favorites. How is this possible? How do you do it? How does it work?
Well, basically, we’re using machine learning technology. And machine learning, basically, the idea is you’re teaching computers to recognize particular patterns in data. And so, if we talk about, for example, the stage favorites, we feed the model, all of the race results we’ve used the last six years of race history, not just for the Tour de France, but for every UCI event from 2.1’s, Hors category, World Tour races, et cetera. Using all of that information, we build up profiles of riders.
So you can start to look at, based on their results historically, you can start identify the profile or the nature of a rider. Which types of courses are they well-suited to, what types of races, how they rank versus other riders, the details of the course that given day. Their team, so, who are the other riders in their team, and how that might influence a result. Generally a stronger team tends to do better.
All of these things are all factors, some of them bigger factors than others, but they all feed into a model. And basically, that model, if I just talk about the stage favorites, it’s about 150,000 different decision points that it’s making.
It’s taking all this time information and it’s taking it through all of these different decision trees and coming out with an output, and basically saying “This is the probability for each rider of winning today’s stage.” It ends up predicting riders, or groups of riders, who are actually pretty much similar to, the ASO run a poll of all the journalists each day, who are the riders that the journalists think are going to be the favorites for today, and inevitably the riders that our model is predicting look very, very similar to the group of riders that the journalists are predicting.
What we’ve demonstrated, I think, with the Tour, is that the sorts of predictions we’re making are similar to the predictions that the experts are making. If you can use technology to make predictions that are similar to highly, potentially expensive, rare experts, you can potentially make that knowledge and expertise available to a much broader range of people, and a much broader range of situations.
Picking The Winner
So, this is starting to raise another big question, another interesting question. A couple of ones. First of all, did you predict Egan Bernal as the winner of the Tour this year?
Yes. He was number one. We made that prediction before the Tour started.
Ah, okay. Wow.
That was primarily because if you actually look at this results and his form leading up to the Tour, he was probably the most in form GC contender in terms of his results, particularly early in the season. So, an interesting thing is that Geraint Thomas the machine rated quite low in terms of his probability of winning, and I think a large factor in that was that this season so far, he hasn’t actually really finished a race, he hasn’t had a GC result. So, the machine actually said, basically the machine’s figuring “Well, he hasn’t had a GC result so therefore his probability of doing well in this year’s Tour is much lower,” because historically, the winners of the Tour de France have typically done well at a race like the Dauphiné or the Tour of Suisse, those earlier season stage races.
You Can Bet On This
The second thing that comes to my mind is if you have the ability to start predicting these outcomes, there’s something much bigger going on here in the world of betting.
Yeah it’s certainly a question that comes up pretty regularly, and those guys are also producing their own models for identifying likely winners, but they’re actually as much interested in where are the betting dollars going, because they’re basically hedging. So, they’re actually more interested in understanding where the dollars are going to go then necessarily who’s actually going to be the winner of the stage. So, the way that you’re actually creating that model is slightly different.
This is great, that we can watch the Tour and get all this data on our screen and play along at home. What is the future for this kind of data collection in terms of sport? What are you as a brand and as a company looking at, and what do you personally think is the future of this?
We see that technology will keep evolving, and the way that we engage with technology will keep evolving. If we look at how we watch sport today versus how we watched it 10 years ago, it’s quite significantly different, because we all have these devices that give us access to a whole bunch of additional information. And so, we know, and we did some research just prior to the tour, we know that a large percentage of people at a live sporting event will be looking at stats and additional information on their second screen. They’ll also potentially be streaming replays or looking at video content while they’re actually at the live event, let alone for people for at home who will be using a digital platform as their second screen, but also then interacting through social media. It’s the digital version of having your mates around to watch the Sunday afternoon footy game.
That technology’s going to continue to evolve, and some of the sorts of trends that we see are the way that we engage with video and television is changing. Moving away from traditional television broadcast to fully live streamed, we’ve got move to higher definition broadcast, there’s some experimentation happening at the moment with three-dimensional views of television, through things like holograms and those types of things. And so, I think there are ways that we’ll see that continue to change and evolve and sport will move with that, and in some ways, sport will probably be one of the leaders in those things. It’s often large sporting events that really bring to the public consciousness new innovations or things like high-definition television. It’s typically been major sporting events that have meant that people go out and invest in their new TV, because the Olympics are coming up, or the World Cup soccer is coming up and so therefore, “I’ll invest in that next step of technology.”
We’re keeping a very close eye on how those things evolve, but then also thinking very carefully about, a lot of the work we’ve done so far is around providing data to fans who are watching the broadcast, particularly watching at home. But how can we change the experience of the fans at the roadside? More and more, if we look at sports like American football or basketball or baseball, the stadium experience is changing, evolving quite substantially as they create these smart stadiums that mean that you’ve got a highly connected environment that you can do things like in-seat ordering of food, get very personalized information, et cetera.
Imagine if you were actually able to create a stadium that’s 3,000 kilometers long around France. And you’ve been to the Tour, you know what sort of chaos that environment is. The irony is, often at the race, it’s the most difficult place to actually find out information about what’s in your face.
Yeah, that and parking.
Yeah. So, imagine enhancing the experience of the fan by the roadside by A, knowing what’s going on in the race, and B, being able to find parking, and then find your way from your parking.
You add to that things like finding merchandise. You want to go buy a Tour de France t-shirt, and trying to find the merchandise stand is not always an easy thing. How do you change that experience? Can I, on my phone, order the t-shirt and maybe it’s delivered to me by the tour caravan as it comes along the roadside. There’s all kinds of potential opportunities in that space around changing the experience of the fans who are there.
Beyond Sport – What’s Next?
What about applications for this kind of technology beyond sport? Do you have any sense of where this might work into a broader sense of our daily lives?
Yeah, so all of the stuff that we’re building and developing with the ASO for the Tour de France is technology that we believe is relevant for our other customers. So, NTT, most of our business is with large corporations around the world. It might be delivering healthcare, it might be delivering logistics services, retail services, for government, for financial services, et cetera. And so, the sorts of things that our customers who are looking at this then become interested in is the applicability of the internet of everything, so if I take the example of healthcare, creating connected spaces within aged care facilities, for example. This is one that we’ve been working on where it allows residents in aged care facilities a better balance between having a controlled environment, but having freedom and flexibility in the way that they live their lives, so it can provide higher quality care and ensure that they’re being really well looked after.
Are you using this technology in other sports beyond cycling?
Yeah, so we’ve taken the same sorts of technology and applied it to, certainly, well, within a number of other cycling disciplines including mountain bike, track cycling, et cetera. But then also iron man marathon and those types of events. And then, NTT is involved in a whole range of different sporting events around the world. So, we’re involved in IndyCar in the US, we’re involved in European soccer and provide services into European soccer. Around the world, we’ve done a number of rugby world cups, cricket world cups, et cetera, in different ways, providing different types of technology services. The Volvo Ocean Yacht Race last year, we were providing services at a number of the ports there. So, we are involved in a whole range of different sporting events, sometimes using technology like we do at the tour, but also delivering other types of technology solutions for different kinds of events around the world.
Well, it sounds like a pretty bright future ahead for us. Thanks very much Peter, pleasure talking to you.
See more at the these links:
• YouTube: https://www.youtube.com/channel/UCoXa9yrN39N2a916TkVNlXw
• @leTourdata on Twitter: https://twitter.com/letourdata
• Twitter: https://twitter.com/GlobalNTT
What to type in: @GlobalNTT
• LinkedIn: https://www.linkedin.com/company/global-ntt/
What to type in: @NTT Ltd. (first result)
• Facebook: https://www.facebook.com/GlobalNTTLtd/
What to type in: @GlobalNTTLtd (first result)
• Instagram: https://www.instagram.com/globalntt/
What to type in: @globalntt