Innovation
How Formula 1 teams determine the fastest race strategy
by Gemma Hatton
7min read
Formula 1 teams run billions of strategy simulations every race weekend to help decipher the fastest way to the finish line. This arms strategists with the intel they need to make those all-important, split-second decisions that can be the difference between winning and losing.
Career
Formula 1 careers: How I made it to F1 – and what it’s like working in motorsport
But how do teams actually go about defining the optimum race strategy? What is the point of all these simulations?
Former Pirelli tyre engineer and F1 strategy engineer Gemma Hatton explores the world of race strategy and the tactics behind winning races.
The role of a race strategist
Unlike the rest of the engineers at the track, race strategists are not directly involved in improving car performance. Instead, they take a more global view and focus on maximising the results of the weekend with the car performance they have available.
To achieve this, they spend the same amount of time analysing the competitiveness of their rivals as they do themselves. This data builds an overall picture of how a race is likely to pan out and is updated live, allowing strategists to quickly respond to track conditions and competitor decisions.
Tyre stints and strategies from the 2024 Austrian Grand Prix
Simulating strategy
There are essentially three main elements at the core of a strategy simulation:
1) Tyre model
2) Free air optimisation
3) Monte Carlo simulation
"Typically, we start off by considering the fundamentals roughly a couple of weeks prior to an event," explains Randeep Singh, Racing Director at McLaren. "There are two fundamental factors that give you a baseline for the right strategy: tyre behaviour and pit loss. The worse the tyre behaviour, the more stops you want to make, and the higher the pit loss, the fewer the stops you want to make."
"From those two fundamental factors you can determine the quickest strategy if you were racing alone," continues Singh. However, the interactions with other cars such as overtaking or following makes the problem much harder and this is where considerations of other cars strategies and game theory come into play."
Tyre models
Tyres have the greatest influence on race strategy in F1. Understanding how each car interacts with its tyres allows teams to gauge the pace of their competitors, and therefore how likely they are to overtake.
Consequently, these tyre models are extremely complex and are based on hundreds of parameters that are continuously updated live during a race.
However, to build a simplistic tyre model, you only need three pieces of information:
1) Base pace – how fast a new tyre is
2) Tyre degradation – how much tyre performance decreases per lap
3) Tyre life – how long the tyre will last
This information can be used to generate tyre curves that model the tyre performance per lap, as shown in the example below.
Example linear tyre degradation curves based on hypothetical data
In this case, the medium tyre’s base pace is 0.7s slower than the soft as highlighted by point A. The soft tyre experiences slightly higher degradation per lap than the medium, as illustrated by the steeper red line at point B. This degradation significantly increases after lap 15 at point C, suggesting that the soft’s tyre life is 15 laps, while the medium’s tyre life is 25 laps at point D.
Free air optimisation
Once the tyre models have been defined, the next stage is to understand the theoretical fastest strategy with the tyres available. To do this, teams conduct a free air optimisation which models only one car on track and assumes constant car performance and fuel consumption. Normalising these factors allows different strategies to be easily compared.
Let’s consider a 55-lap race where the soft and medium are the allocated compounds. The tyre life calculated above dictates this is a two-stop race meaning there are three stints. Consequently, we can either use one soft and two medium sets of tyres, or two softs and one medium.
To work out which strategy is fastest we can plot the time loss per lap of each compound against the average lap time for each stint. We then add in the time spent during pitstops and we end up with the graph below.
Example free air optimisation based on hypothetical data
Both strategies start with the faster soft tyre and gain on the average lap time as shown by the increasing curve at point A. On lap 15 at point B, both strategies complete a pitstop and lose 22 seconds in the pits.
During the second stint, the green strategy opts for another set of softs and again goes faster than the average lap time, until the second pitstop on lap 30, point C. Meanwhile the blue strategy switches to the medium and the slower pace of this tyre makes this stint slower than the green strategy, but last longer until lap 35 at point D.
For the third stint both strategies go for the medium, but because the blue strategy is already slower, it never catches up to the green strategy before the end of the race. Consequently, the green strategy is fastest.
The combination of two fast tyres and one slow tyre is always going to be faster than one fast tyre and two slow tyres. However, this is a very simplistic example based on linear tyre models. In reality, tyre degradation increases at faster rates towards the end of a stint, and so the theoretical fastest strategy is not quite so easy to define.
The Monte Carlo method
Now it’s time to model the influence of other cars on the race strategy. To do this accurately requires adding many more parameters to build a clearer picture of how the race will likely pan out. Alongside higher fidelity tyre models, teams will also incorporate overtaking models, competitor performance metrics, the effect of safety cars, fuel saving and much more into their strategy simulations.
They will then use a statistical analysis method called ‘Monte Carlo’ to run a simulated race. This essentially identifies all the possible outcomes of a particular race scenario and the probability of each of these outcomes occurring. It is then the job of the strategist to vary the inputs and run these simulations again to compare results and understand which changes in inputs cause different results.
The race planner view in RaceWatch. Left of the vertical line at lap 14 are the actual positions of the drivers, while on the right are the predicted outcomes. Image courtesy of Catapult.
"We run simulations that sweep across different tyre behaviours, pit losses, strategies for competitors and ourselves that use a quasi-Monte Carlo method to give us more information about our baseline strategies and about how they may vary," highlights Singh.
"We update this effectively any time we get new data. So before running, updates usually relate to changes in weather or Pirelli prescriptions. During the race we calculate live tyre degradation, overtaking difficulty and our relative pace. We also take into account live wear information once tyres have been used and changes in ambient conditions."
A simulation is only as accurate as its models and will always differ from reality. Furthermore, they only give the probabilities of particular outcomes; not definitive results.
That’s why strategy simulations only form part of the decision-making process. The rest relies on the intuition and experience of the strategy department – and of course, how the drivers and pitcrew execute the strategy.