📊 statistics

The Persistence of the Dirty Air Penalty: A Causal Analysis of Formula 1's 2022 Ground Effect Regulations

This study utilizes extensive lap data and advanced causal inference methods to demonstrate that the 2022 Formula 1 ground effect regulations failed to produce a statistically significant reduction in the "dirty air" penalty, as other racing factors appear to swamp any potential aerodynamic improvements.

Original authors: Bhavay Joshi

Published 2026-08-05

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Bhavay Joshi

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are watching a high-speed race where the cars aren't just fighting each other, but also fighting the air itself. In the world of Formula 1, air isn't just empty space; it's a heavy, invisible ocean that the cars must push through. When a car zooms along, it leaves behind a turbulent, messy wake of air, like the choppy water left behind a speedboat. If a second car tries to follow too closely, it has to drive through this messy wake. This "dirty air" is like trying to run through a crowd of people waving giant, flapping blankets; it strips away the second car's grip and speed, making it incredibly hard to catch up and pass. For decades, this was the sport's biggest headache: the leader was safe, and the chaser was stuck.

To fix this, the sport's rule-makers decided to change the rules of the game in 2022. They wanted to redesign the cars so that the messy wake behind them would be less harmful to the car chasing them. Think of it like changing a car's exhaust pipe so that instead of blowing smoke directly into the face of the car behind, it blows the smoke harmlessly to the side. The big question was: Did this new design actually work? Did the cars behind finally get a fair shot at passing, or was the "dirty air" still a giant, invisible wall? This paper dives into the actual race data to see if the new rules changed the game, or if the cars were still stuck in the same old traffic jam.

The Great Air Experiment: Did the New Rules Work?

In 2022, the Formula 1 rulebook got a massive makeover. The goal was simple but ambitious: stop the "dirty air" penalty. You know that feeling when you're running behind a friend, and suddenly you feel like you're running through molasses? That's what happens to a Formula 1 car when it follows another too closely. The air in front of the leader gets all scrambled, and the follower loses its grip and speed. The new rules tried to fix this by changing how the cars generate downforce (the force that pushes them onto the track), hoping to make the air behind them cleaner and friendlier.

But here is the twist: while the engineers had fancy computer simulations that said, "Yes! This will work!", no one had actually checked the real race results to see if it held up in the messy, chaotic reality of a Grand Prix. That's where this paper comes in. The author, a curious data detective, decided to skip the computer models and look at the raw numbers from over 100,000 racing laps between 2021 and 2025.

The Detective Work: Cleaning the Data
Before solving the mystery, the detective had to clean up the crime scene. Race lap times are messy. A car gets slower as it burns fuel (like a backpack getting lighter, but wait—actually, a car gets faster as it burns fuel, so a heavier car is slower at the start). A car also gets slower as its tires wear out, like running on flat shoes. To find the "dirty air" effect, the author had to strip away these other factors. They built a special mathematical filter to remove the "fuel weight" and "tire wear" noise, leaving behind only the pure speed of the car. This ensured that if a car was slower, it was actually because of the air, not because it was running on old tires.

The Big Reveal: The Penalty Stays
After crunching the numbers with two different super-smart statistical tools (one called OLS regression and another called Causal Forest, which is like a team of decision-making trees), the result was surprisingly quiet.

The paper found no statistically significant evidence that the 2022 rules reduced the dirty air penalty. In plain English: even with the new car designs, the car chasing another car still lost about the same amount of speed as it did before. The "penalty" for following closely didn't go away.

The author looked at the data in two ways:

The Big Picture: They compared all the years after 2022 against the years before. The result? The penalty was still there. The math showed a tiny, almost invisible change that was statistically the same as zero.
Year-by-Year: Maybe the rules worked for a little while and then faded? The author checked 2022, 2023, 2024, and 2025 individually. The result? Still no clear improvement. Some years looked slightly different, but the pattern was messy and didn't show a steady, successful trend.

Why the Computer and the Track Disagree
This is where it gets interesting. The paper acknowledges that the engineers' computer simulations did show a big improvement. In the virtual world, the new cars created cleaner air. But in the real world, the simulation didn't match the race track.

Why? The author suggests that real racing is a chaotic soup of other factors. In a simulation, you can control everything perfectly. But in a real race, drivers are fighting for position, tires are wearing out in weird ways, and fuel loads are changing every second. The paper argues that these real-world factors are so strong that they "swamp" (or drown out) any small aerodynamic advantage the new cars might have had. It's like having a super-fast engine in a car, but if you're stuck in a traffic jam, the engine doesn't matter. The "dirty air" might be slightly cleaner in theory, but on the track, the other problems are still too loud to hear the improvement.

The Bottom Line
So, did the 2022 rules fix the overtaking problem? According to this deep dive into the data, no. The "dirty air" penalty persisted. The paper doesn't say the new cars are bad or that the physics are wrong; it just says that the promised aerodynamic magic didn't translate into a measurable speed advantage in the actual races. The gap between the clean, perfect world of computer simulations and the messy, tire-grinding reality of the racetrack is real, and in this case, the reality won. The cars are still fighting the air, and the chasers are still finding it just as hard to catch the leaders as they were before the big redesign.

Technical Summary: The Persistence of the Dirty Air Penalty in Formula 1's 2022 Ground Effect Regulations

Problem Statement
The 2022 Formula 1 season introduced a major regulatory overhaul shifting the primary source of downforce from wings to ground effect underbody tunnels. The FIA's stated objective was to reduce the "Dirty Air" penalty—the aerodynamic performance loss experienced by a following car due to the turbulent wake of a leader—thereby facilitating closer racing and increased overtaking. While this claim is supported by external aerodynamic simulations (CFD), it has not been rigorously verified against actual race timing data. The core problem addressed is whether the theoretical aerodynamic benefits translated into a measurable reduction in lap time penalties for cars following closely in real-world racing conditions, where confounding factors like tyre degradation, fuel load, and driver strategy exist.

Methodology
The study utilizes a dataset of 100,486 racing laps collected via the FastF1 API, covering the 2021 (pre-regulation) and 2022–2025 (post-regulation) seasons. The methodology involves a multi-stage process to isolate the "Dirty Air" signal from noise:

Data Preprocessing and Filtering: The dataset was restricted to "green flag" laps (TrackStatus == 1) on dry compounds (Soft, Medium, Hard) to eliminate variance caused by safety cars, pit stops, and wet weather. Intermediate compound laps were excluded due to systematic bias.
Physics Correction (Isolating the Signal): To isolate the aerodynamic effect of following distance from other variables, a "CorrectedLapTime" variable was constructed. This involved a joint regression model per tyre compound to estimate and subtract the effects of:
- Fuel Load: Proxyied by lap number within a stint.
- Tyre Degradation: Estimated via TyreLife (laps on current set).
- The model jointly fitted LapTime = α + β₁·TyreLife + β₂·LapNumber + ε to prevent collinearity between fuel burn and tyre aging.
Causal Inference Models: Two independent statistical techniques were employed to test the hypothesis that the 2022 regulations reduced the penalty:
- OLS Regression with Clustered Standard Errors: A linear model testing the interaction between GapToCarAhead (time gap to the car ahead) and IsPost2022. Standard errors were clustered by race to account for serial correlation within events. A strict circuit filter was applied to include only circuits present in both eras to avoid collinearity between circuit and era fixed effects.
- Causal Forest Double/Debiased Machine Learning (DML): A non-parametric approach used to detect heterogeneous effects and non-linearities that OLS might miss. This model treated GapToCarAhead as the treatment and CorrectedLapTime as the outcome, controlling for circuit, compound, tyre life, and lap number.
Robustness Checks: Both models were re-run with "Season" as a categorical variable (2021 baseline vs. 2022–2025 individually) to test for a specific alternative hypothesis: that any benefit was achieved in 2022 but eroded in subsequent years.

Key Contributions

Reproducible Dataset: Creation of an open-source, lap-by-lap dataset spanning five seasons with a specific methodology for physics-based correction of lap times.
Dual-Method Validation: Application of both traditional econometric (OLS) and modern machine learning (Causal Forest DML) techniques to validate findings against potential non-linearity and effect heterogeneity.
Decoupling Simulation from Reality: A direct empirical test separating FIA's simulation-based engineering claims from actual on-track performance outcomes.

Results
The analysis yielded consistent null results across both methodologies and temporal granularities:

Pooled Analysis: Neither the OLS interaction term nor the Causal Forest mean effects showed a statistically significant decrease in the Dirty Air penalty post-2022.
- OLS Interaction ( $\beta_3$ ): $-0.0025 $($ p = 0.803$).
- Causal Forest Mean Effects: 0.019s (2021) vs. 0.018s (2022–2025) per second of following distance. The distributions of treatment effects were nearly identical and overlapping.
Season-Level Analysis: When broken down by year, no individual season (2022–2025) showed a statistically significant difference from the 2021 baseline in the OLS model.
- While Causal Forest estimates showed some non-zero effects in specific years (e.g., a negative effect in 2025), the pattern was non-monotonic and did not align with the "rise and fade" erosion shape suggested by simulation data.
Descriptive Findings: Corrected lap times showed a non-monotonic, U-shaped relationship with following distance (faster in the 0.5–1.0s bin than in the 0–0.5s bin), suggesting driver behavior (e.g., pushing harder in mid-range gaps) may influence lap times more than pure aerodynamic wake effects.

Significance and Claims
The paper concludes that there is no statistically significant evidence that the 2022 ground effect regulations reduced the Dirty Air lap time penalty in the analyzed racing data. The author emphasizes a critical distinction between simulation and reality: even if an aerodynamic advantage exists in controlled CFD simulations, it may fail to manifest in actual race times due to the "swamping" effects of tyre strategy, driver behavior, and race context.

The study does not refute the existence of the aerodynamic phenomenon in theory but asserts that the claimed performance benefit has not translated into a measurable outcome on the racetrack. The findings suggest that the "Dirty Air" penalty persists under the current regulations, challenging the narrative that the 2022 rules successfully solved the overtaking problem through aerodynamic redesign alone. The paper maintains a modest stance, noting that limitations such as the use of lap number as a fuel proxy and the potential for unobserved confounders in the DML assumption prevent definitive causal claims beyond the scope of the available data.

The Great Air Experiment: Did the New Rules Work?

More like this