F1 Pit Prediction

When to pit is vital to an F1 team’s racing strategy as it can give a team huge advantages but also make them lose time, so it is beneficial to know when other teams pit

Comment

Inspiration

We love to watch F1. Formula One collects massive amounts of information very second, so the data scientists and Formula 1 enthusiasts in us wanted to combine our passions We know that a team's pit timing is vital to their racing strategy as it can give a team huge advantages but also make them lose time, so it is beneficial to know when other teams pit. For example, if a car pits it can either end up behind a slow car or in front a pack which can slow them down or give them open, increasing their speed, respectively.

What it does

Our model predicts if a driver is going to pit on a certain lap given data like their laptime, weather data, and more.

How we built it

We used an LSTM to input our data into the model every time step (every lap) and the output is a probability between (0,1) if a driver is not pitting or pitting respectively. We then compare that probability to the ground truth to compute our loss.

Challenges we ran into

  • Some drivers did not finish the entire race, padded the data to fit LSTM
  • Weather data did not fully align with Driver race times
  • Minimal data on Pit Stop counts, created bin width to increase observation counts
  • The data that we obtained was disorganized and took a long time to figure out what important information there is
  • A lot of data augmentation

Accomplishments that we're proud of

  • Learned about RNN and LSTM in a short period of time
  • We were able to conduct deep EDA and decide what predictors we can create and augment

What we learned

  • We learned how RNN and LSTM's work and how to apply them to data.
  • Working with large amounts of data that was disorganized

What's next for F1 Pit Prediction

  • Attempt to try a poisson process since it theoretically can work better for sparce data.
  • More feat such as delta time distances between 1 driver and all other drivers (20 drivers per race)
  • Watch all of the races to track the racer's tyre compound as that is a vital predictor. Some tires are made of harder material so it wears slower and some tires are made of softer materials so they wear faster.

Built With

Share this project:

Updates