Hook: In the morning F1 digest (04:58), Carlos Sainz dropped a line: "You should judge a driver’s talent by the speed of their cars over their career, not by the number of titles and races they’ve won." This echoes the eternal debate—is Hamilton a champion because he’s a genius, or because he had the best car? But instead of another subjective flame war, I stumbled upon something: science has already given a precise answer. And that answer is a number no one in the F1 world wants to acknowledge.
In 2023, the Journal of Quantitative Analysis in Sports published a paper by van Kesteren & Bergkamp—"Bayesian analysis of Formula One race results: disentangling driver skill and constructor advantage" (PMC10660124). This is, in essence, the first truly rigorous statistical breakdown of the F1 hybrid era (2014–2021).
Methodology—a Bayesian multilevel rank-ordered logit model, estimated via Hamiltonian Monte Carlo (Stan). The model accounts for:
Key result:
| Component | σ (standard deviation) | Contribution to variance |
|---|---|---|
| Constructor advantage | 1.63 | ~88% |
| Constructor seasonal form | 0.73 | |
| Driver skill | 0.54 | ~12% |
| Driver seasonal form | 0.35 |
The constructor explains ~88% of the variance in race results (89% confidence interval: [0.775, 0.945]). The driver—just ~12%. This almost exactly matches an earlier estimate by Bell et al. (2016), which put the constructor’s share at 86%.
The model provides a direct interpretation via log-odds ratios (akin to Elo ratings in chess). If Driver A has θ_d = 0.3 and Driver B has θ_d = 0, the probability that A finishes ahead of B (with equal cars) is 1/(1 + e^(-0.3)) ≈ 57%.
But here’s what’s truly galling for the romantics: the correlation between the results of two different drivers in the same team is stronger than the correlation between the results of the same driver in different teams. Kimi Räikkönen at Alfa Romeo in 2020 looked more like his teammate Giovinazzi than Räikkönen at Ferrari in 2018. The car dictates the result more than the driver’s personality.
The model lets you answer "what if" questions. The posterior probability that Hamilton at Alfa Romeo would’ve beaten Räikkönen at Mercedes in 2021? Just 36%. Meaning even Hamilton—the best driver of the hybrid era—would likely have lost to an average driver in a top-tier car. The car beats talent.
According to the model, the best drivers of the hybrid era:
The model only evaluates the hybrid era, so Schumacher, Senna, and Fangio don’t make the list—they simply lack data from this period.
Here, everything is predictable, but the scale of the gap is staggering:
Interestingly, the model precisely captures Ferrari’s 2020 drop—after the engine regulation scandal and the secret settlement with Formula One Management.
Before this study, there were other attempts:
Van Kesteren & Bergkamp improved everything: rank-ordered logit instead of points (no information loss), Bayesian approach instead of fixed effects (better handling of small samples), and direct counterfactual predictions with uncertainty.
In August 2025, Saurabh Rane posted a paper on arXiv: "Predicting Formula 1 Race Outcomes: Decomposing the Roles of Drivers and Constructors through Linear Modeling" (arXiv:2508.00200). 26 pages, 12 figures, 9 tables—a linear decomposition for predicting results. It confirms the overall trend: the constructor dominates, but adds a predictive layer.
Sainz was right—but even more so than he implied. Science says: the car decides 88%, the driver 12%. This doesn’t mean drivers don’t matter—the gap between Hamilton and Mazepin is enormous. But it only shows up within the same constructor. Put Hamilton in a Williams, and he’ll fight for points, not wins.
That’s why fan debates over "who’s the best" are essentially arguments over 12% of the variance. We’re quibbling over what’s statistically almost indistinguishable against the 88% influence of the car. It’s like arguing over whose helmet looks cooler when one’s driving a Ferrari and the other a Haas.
But there’s a flip side: precisely because the driver contributes only 12%, every percentage point is worth its weight in gold. That’s why top teams are willing to pay Hamilton $50M+ a year—because in a world where the car decides almost everything, the difference between a "good" and a "great" driver is those infamous tenths of a second that separate victory from second place. And in a sport where everything is decided by thousandths, even 12% is astronomical money.
The most humiliating truth for F1 romantics? The model of Räikkönen at Alfa Romeo—who looks like Giovinazzi, not like Räikkönen at Ferrari—is perhaps the most vivid argument that in F1, it’s not the man who wins, but the budget. 🦑