Amazon and F1 deal will turn raw data into racing entertainment
Former Ferrari chief mechanic helping AWS bring added insights to Formula One fans
At the recent Bahrain Grand Prix, AWS’s insights proved eerily prescient, predicting just when Red Bull’s Max Verstappen would be close enough to Mercedes’ Lewis Hamilton to start fighting for the lead of the race
Formula One is many things. It’s a sport. It’s a source of national pride for many. It’s a massive, global, multibillion-dollar industry. What it mostly is, these days though, is a source of data. Colossal amounts of data, terabytes upon terabytes, all generated from sensors and transponders that carefully monitor every part of each racing car, as they whoosh by at 300km/h.
Making sense of that data has, until now, been the preserve of those people within F1 teams. The ones sat behind banks of screens on the pitwall, or down the back of the garage, dressed in brightly coloured team clothing, but staring seemingly forever at Matrix-like readouts of numbers telling them the pressure in every tyre, the fuel being pumped into the engine, the twitch of airflow over each wing, and so much more.
Which can be rather distancing for the average Sunday afternoon viewer. After all, Formula One began as effectively a petrol-fuelled gladiatorial contest, mano-a-mano, death-defying, wheel-to-wheel combat. These days? It’s more of an engineering exercise, a careful honing of every component and widget before some daring young buck gets behind the wheel to drive to glory. So much of the winning is done by the engineer, by the car’s designer.
Now, Amazon Web Services (AWS) has been brought on board by Formula One (the company itself, the one that – as a subsidiary of Liberty Media – actually owns the rights to the sport) to help take all of that data, all of those terabytes, and make sense of them for the fans. Those who regularly watch F1 on the telly will have noticed in the past two seasons little AWS-branded graphics popping up, giving helpful insights into the way the race is playing out. Effectively, it’s a way to give armchair fans a seat on the pitwall. Starting this year, the information being fed to that seat is going to become even richer.
Rob Smedley is the gamekeeper-turned-poacher here. Having started his F1 career with the Irish Jordan team, Smedley eventually moved to Ferrari and became one half of a devastatingly effective double-act – the engineering partner to Brazilian racer Felipe Massa. The pair – and it was a genuine partnership, not just a sidekick role for Smedley – were beaten to the 2008 F1 driver’s title by only a single point.
Now, it’s been Smedley’s job to convince F1’s team bosses to hand over once-closely guarded data to Amazon. I suggest to Smedley that it’s a request which would, in his previous team role, probably have given him a panic attack.
“It’s a philosophical point, I guess,” Smedley tells The Irish Times. “What we’ve tried to get across to the teams is that we’re not using the data for data’s sake, but trying to tell the sporting story that’s happening, and to give that insight to the fan. The teams are clever enough, switched-on enough, that even though they are giving something away, they can see that the effect on the whole of Formula One is much bigger.”
AWS isn’t taking the entire data feed from each car – if it did, rival teams would be able to exploit the data for competitive advantage, and that’s something that Smedley says F1 and Amazon has worked hard to avoid – but it’s getting a lot, and is mixing it with information taken from other sources, such as GPS measurements of the car’s speed and position on track, and of course the actual stopwatch timing that tells you just how fast a given car is lapping at any one time.
That data – within milliseconds – is fed through multiple layers of cloud computing and Amazon’s machine learning SageMaker software. It then beams back, effectively, predictions. Predictions of where one car might find itself on track in a given number of laps, and how that relates to cars with which it’s competing.
At the recent Bahrain Grand Prix, AWS’s insights proved eerily prescient, predicting just when Red Bull’s Max Verstappen would be close enough to Mercedes’ Lewis Hamilton to start fighting for the lead of the race. The on-screen graphics added an exciting countdown clock, enlivening what might otherwise have been a dry tactical battle.
“There’s a balance to be struck,” says Smedley,“a balance between what we can tell people on screen, and what they really want to see – which is drivers fighting wheel-to-wheel on track.”
Unquestionably, there’s also a sense that by giving fans direct onscreen access to the same sort of data that the team engineers are seeing, a certain level of papering over the cracks of an occasional dull race can be achieved.
Smedley is also alive to the fact that too much data, and too much reliance on data, can be self-defeating. F1 fans were up in arms last year when an Amazon run-down of the greatest F1 drivers of all time only used data from 1983 onwards (thereby eliminating the likes of Jim Clark, Juan Manuel Fangio and Stirling Moss) and included the likes of Heikki Kovalinen and Jarno Trulli – solid drivers but hardly world-beating talents – in the top 10. Since then, there seems to be an awareness at Amazon that focusing too heavily on stats takes away from the human side of the sport.
“We’ve got to be constantly listening to the fans,” says Smedley. “Because if we’re putting too much information on screen, it just becomes white noise, it turns people off, and we definitely don’t want that.”
For the 2021 season, AWS and F1 will be rolling out some new onscreen insights. Starting with the Emilia Romagna Grand Prix at Imola in Italy, AWS will be able to tell fans how hard and how late a driver is braking for a given corner, and also how just how hard they are pushing their car relative to its total potential performance.
Vast computing resources
However, Amazon’s partnership with F1 goes beyond mere onscreen graphics. The retail giant has turned its vast computing resources over to Formula One to help it solve the greatest problem of the age: getting the cars to race more closely together.
Because F1 cars use vast wings, generating downforce that pushes the car into the track, making them go faster around corners, they also cause turbulence. That disturbed air is, essentially, poison for any rival car trying to come up behind, making overtaking – the lifeblood of the sport’s competition – close to impossible at some tracks. F1 and AWS have been working together, building a virtual wind tunnel in Amazon’s cloud computers, tying together vast numbers of processors to mimic the actions of a multimillion-euro supercomputer. With that computing power, AWS has been able to help F1 shape both its regulations and the physical bodywork of 2022’s cars, to try to reduce the turbulence and make the racing closer.
“This is the whole beauty of the diverse partnership we have between F1 and AWS,” says Smedley. “You know, you bring in guys like me – who have a history of expertise in vehicle performance and dynamics – and you combine that with machine learning and the computing power of AWS and you kind of build a supergroup. A series of real key skill sets that come together. Then you assess the data, take lots of different ingredients, and come up with a specific output.”
That specific output, as Smedley puts it, is a new generation of racing cars that could, should, hopefully will prove far more competitive with each other, allowing the drivers to really go wheel-to-wheel with one another. If it works, it could be the best Amazon delivery of them all.