A Hidden Markov Analysis of Wet-Weather Tyre Crossover in Formula One
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Statistics. The day in statistics: methods, theory and the inference under them.
Tom: We're looking at the paper: "A Hidden Markov Analysis of Wet-Weather Tyre Crossover in Formula One".
Jane: Here's the gist of the paper. Detecting a wet-weather regime and deciding when to replace slick tyres are distinct tasks This study examines their relationship through a retrospective analysis of the 2024 British Grand Prix The combined rule enters on lap 27, matching the team-weighted modal reference These comparisons illustrate how state inference can contextualize a performance criterion while exposing dependence on model assumptions Data and Observation Setup The analysis utilized 960 car-laps from 19 drivers, deriving lap timing, tyre compounds, pit events, and accuracy indicators from FastF1 The observation window ran from race start to the final recorded precipitation sample Normalized lap time was calculated as qi,t = 100 Li,t / bi, where bi is the median of three fastest clean slick laps completed before the first positive precipitation report The HMM observes xt = (gt, rt)⊤; excluding the opening bin leaves 33 consecutive complete observations The Hidden Markov Model Structure The state model consists of three states: Dry, Damp, and Wet statistical regimes denoted as zt ∈ (D, M, W) The emission model is x˜t zt = k ∼ N (muk, sigmak) The initial distribution and transition matrix are fixed, imposing persistent, adjacent-regime transitions and excluding direct Dry–Wet changes within one step Emission means and full covariance matrices were fitted by expectation–maximization using a modified covariance update in hmmlearn The three-state specification allows an intermediate regime between near-baseline and higher-rain, slower-pace observations Entry Rule Development and Sensitivity The qualifying indicator ut = 1 if rt > 0, κ(W) ≥ τ, gt ≥ γ The signal and effective entry times are defined by Tsignal = min (t: mÖ−1 h=0 ut−h = 1) and Tentry = Tsignal + 1 Matched component comparisons remove either gt ≥ γ or κ(W) ≥ τ, retaining positive precipitation, confirmation length, and next-bin action timing Decision sensitivity crosses seven posterior thresholds from 0.40 to 0.
Title and authors: Tom: So we're talking about this paper's title and authors now; it’s "A Hidden Markov Analysis of Wet-Weather Tyre Crossover in Formula One." The authors are Bo-Yi Lin and Michka Sharpe from Cranfield University.
Jane: They’re looking at how the hidden nature of the weather changes when a driver decides to switch tyres during a race, which is pretty specific compared to just general weather prediction.
Lu: What they do is use this HMM structure to analyze six hundred forty-two car-laps and thirty-three complete lap-bin observations from the two thousand twenty-four British Grand Prix <ref:2611.14538#pg1,642 car-laps and 33 complete lap-bin observations>.
Meng: It’s interesting that they are using team-weighted modal reference for the entry rule, which suggests they are trying to get a consensus on when to switch rather than just one driver's decision.
Lalam: The implication here is that you can use this model to see if switching tyres is really tied to an actual shift in the weather regime or just some other performance signal.
The paper's summary: Tom: To summarize the core finding, they are testing how state inference can provide a context for a performance criterion, which is deciding when to change your tyres.
Jane: They set up this model where the HMM observes two things at each lap: the normalized lap time and the five-minute precipitation accumulation.
Lu: The emission means show that there are distinct parameters for Dry, Damp, and Wet states; for instance, in the Wet state, you see a higher mean normalized lap time of one hundred ten point eight three percent compared to one hundred point twenty seven percent in the Dry state <ref:2611.14538#pg3>.
Meng: And they fit these emission means and full covariance matrices using expectation-maximization on hmmlearn, which gives them the statistical parameters for those regimes.
Lalam: The transition matrix is fixed, meaning it only allows you to move between adjacent states—like from Dry to Damp—but not directly from Dry straight to Wet in one step.
Tom: The key operational part is the entry rule, which requires both a high filtered wet-state probability and sufficient pace deterioration.
The paper's improvements: Jane: Now for the suggested improvements; they are looking at how they can refine their decision making process based on state inference and pace deterioration.
Lu: One major improvement is developing a strategy model that uses this HMM state inference to provide a trigger for tyre changes, so you decide when to consider switching from slick to intermediate tyres.
Meng: That sounds useful for production because it moves the decision out of just reacting to rain and into a more predictive sequence based on the inferred regime.
Tom: They also suggest an entry decision mechanism that combines the filtered HMM state probability with pace deterioration, which is what they show requires both to trigger a crossover.
Lalam: Furthermore, they propose matched component comparisons where you can remove either the pace condition or the state condition to see if one drives the decision more than the other.
Jane: That’s interesting because it helps you assess whether your decision is really driven by knowing if it's wet or just by how much slower things are getting.
Conclusion: Tom: So wrapping up, the paper shows that state inference can be used to contextualize a performance criterion like tyre replacement.
Jane: The main implication is that you don't just react to precipitation; you can use a model like this to understand the underlying weather regime and how that affects pace.
Lu: They demonstrate how changing the transition specification and refitting the emissions materially alters when the entry rule triggers, showing sensitivity to those underlying assumptions about regime persistence.
Meng: For me, what matters is that they show *how* sensitive their decision is to those assumptions, which is a lot for someone trying to deploy something robust.
Lalam: The paper also flags a limitation: the probabilities they calculate are retrospective estimates based on the data available up to that point in the race.
Tom: So, in short, "A Hidden Markov Analysis of Wet-Weather Tyre Crossover in Formula One" gives us a framework for using state inference to inform tyre strategy decisions under varying weather conditions.
Cranfield University · CodeF(x) LLC
stat.AP
Submitted: 2026-10-14
Updated: 2026-10-14
Code: https://github.com/theOehrly/Fast-F1
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 79/100
The gist: Detecting a wet-weather regime and deciding when to replace slick tyres are distinct tasks This study examines their relationship through a retrospective analysis of the 2024 British Grand Prix The…
Key concepts
- Hidden Markov Model (HMM)
- An HMM is a statistical tool used to model systems that are in one of several unobservable states, like the weather on a racetrack. The model uses observed data, such as lap times and precipitation reports, to infer which underlying state (Dry, Damp, or Wet) the car was actually in at any given moment.
- State Inference
- This is the process of using historical data and mathematical models (the HMM) to guess the current unobservable condition of a race. Instead of just looking at whether it is raining now, state inference helps determine if the track is truly 'Wet' or just 'Damp,' which leads to more accurate predictions for tyre strategy.
- Matched Component Comparisons
- This technique involves comparing different ways to define an entry rule. The study compared rules based on precipitation alone versus those based on a combination of rain probability and pace loss. This comparison helps determine if the decision is driven by the inferred weather state or by how much slower the car is performing.
- Transition Specification Sensitivity
- This refers to how much the model's assumptions about how quickly a track changes from one weather state to another affect the final decision. By testing different rules for these transitions, researchers found that changing these assumptions significantly altered when the system triggered a tyre change.
Terminology
Summary
Detecting a wet-weather regime and deciding when to replace slick tyres are distinct tasks This study examines their relationship through a retrospective analysis of the 2024 British Grand Prix The combined rule enters on lap 27, matching the team-weighted modal reference These comparisons illustrate how state inference can contextualize a performance criterion while exposing dependence on model assumptions
Data and Observation Setup
The analysis utilized 960 car-laps from 19 drivers, deriving lap timing, tyre compounds, pit events, and accuracy indicators from FastF1 The observation window ran from race start to the final recorded precipitation sample Normalized lap time was calculated as qi,t = 100 Li,t / bi, where bi is the median of three fastest clean slick laps completed before the first positive precipitation report The HMM observes xt = (gt, rt)⊤; excluding the opening bin leaves 33 consecutive complete observations
The Hidden Markov Model Structure
The state model consists of three states: Dry, Damp, and Wet statistical regimes denoted as zt ∈ (D, M, W) The emission model is x˜t zt = k ∼ N (muk, sigmak) The initial distribution and transition matrix are fixed, imposing persistent, adjacent-regime transitions and excluding direct Dry–Wet changes within one step Emission means and full covariance matrices were fitted by expectation–maximization using a modified covariance update in hmmlearn The three-state specification allows an intermediate regime between near-baseline and higher-rain, slower-pace observations
Entry Rule Development and Sensitivity
The qualifying indicator ut = 1 if rt > 0, κ(W) ≥ τ, gt ≥ γ The signal and effective entry times are defined by Tsignal = min (t: mÖ−1 h=0 ut−h = 1) and Tentry = Tsignal + 1 Matched component comparisons remove either gt ≥ γ or κ(W) ≥ τ, retaining positive precipitation, confirmation length, and next-bin action timing Decision sensitivity crosses seven posterior thresholds from 0.40 to 0.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Develop a strategy model that uses HMM state inference to provide a trigger for tyre changes, as
The entry rule could supply a trigger for a strategy model.
This system can decide when to consider switching from slick to intermediate tyres based on the inferred wet-weather regime and pace deterioration. -
Implement an entry decision mechanism that combines filtered HMM state probability and pace deterioration, as the study shows
an entry rule requires both a high filtered Wet-state probability and sufficient pace deterioration.
This allows for more nuanced tyre crossover decisions than relying on precipitation alone. -
Create a system capable of performing matched component comparisons to assess the independence of conditions, by using "Matched component comparisons remove either gt >= gamma (posterior-only) or alpha(W) >= tau (pace-only)." This helps determine if the decision is driven by state inference or pace deterioration when both are present.
-
Build a system that uses transition specification sensitivity to adjust its decision threshold dynamically, as the paper demonstrates that
Changing the transition specification and refitting the emissions materially altered the entry.
The system can adapt its entry criteria based on how sensitive it is to assumptions about regime persistence. -
Design a prospective evaluation framework using
Independentevent evaluation and a model of tyre-change consequences are required to assess prospective strategy value
to test whether an action improves race outcomes, moving beyond retrospective correlation.
Abstract
Detecting a wet-weather regime and deciding when to replace slick tyres are distinct tasks. This study examines their relationship through a retrospective analysis of the 2024 British Grand Prix. A three-state hidden Markov model (HMM) represents normalized lap time and precipitation, while an entry rule requires both a high filtered Wet-state probability and sufficient pace deterioration. The analysis includes 642 car-laps and 33 complete lap-bin observations within the available precipitation record. Each completed-lap signal takes effect on the following lap. The combined rule enters on lap 27, matching the team-weighted modal reference; removing the pace or state condition advances entry to laps 25 and 20, respectively. The combined entry is retained in all 10 team-exclusion refits, but refitting a specification with larger off-diagonal transition probabilities produces entry on lap 20. These comparisons illustrate how state inference can contextualize a performance criterion while exposing dependence on model assumptions. Because estimation and rule development use the same event, the correspondence establishes neither prospective accuracy nor a race-time advantage.
Related papers
- Unraveling time-varying causal effects of multiple exposures: integrating Functional Data Analysis with Multivariable Mendelian Randomization
- Strategic Causal Policy Learning with Manipulable Eligibility
- BayesJudge: Uncertainty-Aware Bayesian Meta-Evaluation of Human and LLM Judgments
- Automating structural reliability analysis with a multi-agent large language model framework
- Improving Minority Population Sampling with BISG Probabilities: Evidence from a Survey of Jewish Americans
- When Do Sports Leagues Stop Being Competitive? Bayesian Detection of Competitive-Balance Eras from Match-Level Data