F1 insights by AWS: ruining the action or enhancing the experience?

Innovation never stops in Formula One - all the way down to the viewer broadcast experience. Formula One’s collaboration with Amazon Web Services (AWS) served to renovate the viewer experience through the use of AI insights - but how exactly can AI models predict the chance of an overtake or the condition of a set of tyres?

To answer the question you never needed to know the answer to - how many data points does a modern F1 car transmit per second? 1.1 million. That’s a lot of data. What AWS aims to do is feed some of this telemetry (tracking) data into a machine learning algorithm called SageMaker, where the algorithm compares the data to historical data to provide predictions that are used to power the F1 Insights broadcast graphics. These predictions can be used to estimate tyre performance, likelihood of an overtake and even race strategy.


The video above shows how these broadcast graphics are integrated into a real-life racing scenario; this was from the 2019 German Grand Prix - one of the most unpredictable races in history. This Battle Forecast graphic feeds lap time data into the SageMaker algorithm (in this case the lap time data of Leclerc and Verstappen) as well as accounting for on-track history between the battling drivers and overall car performance to determine a rough estimate of the difficulty of a potential overtake. For example, if historical data shows that Leclerc has often been unsuccessful in overtaking Verstappen, overtake difficulty would be high. Moreover, if the telemetry data showed that Leclerc’s tyres are of much better condition then Verstappen’s, overtake difficulty would be lower. Thus, SageMaker predictions utilise a whole host of data from various different factors to keep the viewer up to date on possible on-track duels.

However, this isn’t to say that SageMaker is accurate all the time. The F1 insights graphic came under much backlash at the 2019 Japanese Grand Prix in Suzuka after the official tyre supplier, Pirelli, called the graphic ‘misleading’ and ‘inaccurate’. This came after the SageMaker algorithm predicted that Lewis Hamilton’s tyre condition was around 40%, whilst the car ahead’s tyre condition was 80%. Therefore many viewers were left surprised when Hamilton caught up to the car ahead and nearly overtook him, given the AI models implied that he should be much slower than the car ahead.

In conclusion, AI’s use in F1 broadcasting to enhance the viewer experience is a mixed bag. It brings a level of insight and analysis that many viewers aren’t accustomed to and uses historical and live data that would otherwise never see the light of day. However, the graphics aren’t fully accurate and some graphics - especially those predicting the likelihood of an overtake - are criticised to be ruining the unpredictability of Formula One. Thus, it’ll be interesting to see, once the 2020 season gets underway, how F1 chooses to utilise AI in their broadcasting to best enhance the viewing experience, without ruining the action.

Thumbnail from RaceFans and YouTube video from AWS themselves