Bayesian Uncertainty Quantification for Formula 1 Racing Strategies

About the Session

In Formula 1, the fastest car doesn't always win, but having the most statistically robust strategy maximises the probability of success. This session demonstrates how to build a Bayesian race strategy engine using PyMC and FastF1 data, shifting from simple point estimates to Uncertainty Quantification (UQ).

We will focus on the technical challenge of "updating" our models throughout a race weekend. By treating tire degradation and track evolution as probability distributions, we can move from Friday’s practice laps to Sunday’s pit-wall decisions with mathematical precision. We’ll explore how to refine our model's "beliefs" in real-time as new data points roll in from Qualifying and the race itself.

Technical highlights include:
* Sequential Updating: Using free practice "long runs" to inform priors for race-day stint simulations.
* Hierarchical Regression: Modeling tire-wear "drop-off" curves while accounting for epistemic uncertainty in track temperature and fuel loads.
* Bayesian Decision Theory: Moving beyond "best-case" scenarios to Expected Utility, calculating optimal pit windows under traffic constraints.

This code-centric talk is for data scientists and engineers interested in probabilistic programming and decision-making under pressure.

Wesley Boelrijk

Lead Data Scientist,
KLM Royal Dutch Airlines

About the Speaker

Wesley is a Lead/Senior Data Scientist at KLM Royal Dutch Airlines, where he develops next-generation demand forecasting models with a strong focus on time series. Previously, he led a Machine Learning traineeship at Xccelerated (Xebia), mentoring engineers and scientists through tutoring and live coding, alongside various consultancy projects.

A three-time PyData speaker, Wesley is passionate about Bayesian modeling and making forecasting techniques practical. He holds master’s degrees in Econometrics (VU Amsterdam) and Engineering & Policy Analysis (TU Delft), and a bachelor’s in Aviation Engineering.
Outside of work, he is a devoted Formula 1 fan, fascinated by the sport’s rich data streams and the opportunities they create for advanced modeling. He has attended Grands Prix in Zandvoort, Spa, and Spielberg.

Scroll to top