EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
3175. Fast Laps and Fast Stats: A Scalable Spatial Model of Circuit Topographies and their Performance Impact for Constructor and Drivers in F1
Invited abstract in session WB-16: Sports analytics, stream OR in Sports.
Wednesday, 10:30-12:00Room: 19 (building: 116)
Authors (first author is the speaker)
| 1. | Stephanie Kovalchik
|
| Zelus Analytics |
Abstract
Understanding how a car and driver will perform on a given circuit is a central concern of constructors in motorsport. Statistical modelling can be a powerful tool for analysing the interplay between car, driver and circuit features through partial pooling across fine-grained circuit topographies. Telemetry data provides a rich set of inputs for spatial models, yet, the size of the data requires highly scalable methods. We have developed a framework to approximate a Bayesian hierarchical Gaussian process that allows the estimation of latent circuit topographies and their interaction with car and driver at scale. Our fast approach to modelling the spatial characteristics of circuits is applied to the performance outcomes for all races of the 2023 Formula 1 season. Our results demonstrate the predictive and descriptive utility of the framework as it can predict results for new circuits within season and find topographical features where car and/or driver are expected to have the greatest and least advantage over the field.
Keywords
- OR in Sports
- Stochastic Models
Status: accepted
Back to the list of papers