Bachelor's thesis - Maastricht University - 2026
Graded 9 / 10
Data-Driven Evaluation of Strategy and Performance in Formula 1
Quantifying upgrade effectiveness, race-strategy modelling, and the impact of the cost cap and logistical constraints
The capstone of a BSc Business Engineering degree, researched during an internship at Pole Position Reports under journalist Inga Stracke, with Yangzhuoran Yang and Renaud Jolivet as advisors at Maastricht University. It evaluates strategy and performance in Formula 1 from the data the sport produces: the same lens every analysis here is written through, taken to the depth only a thesis allows.
Abstract
Formula 1 is a sport where on track performances are defined by the hard work that goes on behind the scenes, including engineering processes, driver preparation, and transportation processes of travelling the world.
This research identifies a gap in treating these elements collectively. Prior research handles them in isolation, disregarding the notable effect of the cost cap influencing every and all decision made by Formula 1 teams. The methods used include public timing data through a shared lap level pipeline, predictive models for strategy, difference-in-differences for upgrades, and qualitative analysis for constraints.
Findings show that lap performance is highly predictable (validation MAE of 0.78s, under 1% of lap time), and strategy timing is bounded (top 3 hit rate 0.50) by safety car unpredictability. Furthermore, upgrades entail difficulties in quantification only deeming 4 out of 12 significant within the investigated upgrade packages, while development is the binding constraint in resource allocation under the cost cap.
This bounded but feasible strategy aid bodes well as a supportive model for live strategy decisions, while qualitative and partial quantitative methods clearly indicate how tackling upgrade and development decisions within the known cost and logistical constraints is critical to extract maximum performance on track.
The question
To what extent can data-driven methods quantify the impact of Formula 1 car upgrades and race strategies on on-track performance under cost cap and logistics constraints?
- How reliably can tyre degradation and stint length be modelled to predict pit stop windows and strategy choices?
- How effectively can a combined predictive and simulation-based policy recommend race strategy decisions?
- Do in-season aerodynamic and mechanical upgrades result in measurable on-track performance gains, as evidenced by a difference-in-differences (DiD) analysis?
- What do cost cap and logistics constraints reveal about how teams prioritise development and operational decisions?
What it found
- Lap performance is highly predictable: a 0.78 s average error on the unseen 2025 season, under 1% of a lap.
- The strategy policy is a useful but bounded aid: the real pit lap is on its three-lap shortlist about half the time, capped by safety cars no lap data can foresee.
- Simpler models travelled better between seasons: a regularised logistic regression beat gradient boosting, 0.503 to 0.297 on the top-three hit rate.
- Upgrades are only partly measurable: 4 of 12 packages were significant at 95%, led by McLaren's 2023 B-spec at +0.86 s against the field.
- Under the cost cap, development is the binding constraint; in-race strategy is close to costless next to it.