The 2019 F1 Aero Regulations- did they work?
- •
- 531 reads
- •
- 702 views
- •
- 12 min read
As readers may remember, the 2019 F1 aerodynamic regulations were made public in 2018, with great fanfare. They were intended as a ‘quick win’ from F1’s technical research programme, which was targeted mainly at the grand revolution we’ll all see in 2021.
That quick win was supposed to be an interim step to improving the chance of an overtake once two cars are close, albeit to a lesser degree than is expected in 2021.
With the hype surrounding the now-published 2021 regulations, it’s easy to forget what happened in 2019. There has been the occasional, usually not too positive, driver interview about the changes, but little in terms of objective analysis.
Why should F1 fans be interested in what is now consigned to history? Well, F1 is obviously on a major journey heading into 2021 and the only hard evidence the public has that it’s actually moving in the right direction is the success (or otherwise) of the 2019 season in meeting its stated objective of improving overtaking, even if only moderately.
If you want to know the answer, read on; it’s a fascinating story…
How can we measure the effectiveness of the regulations?
Sounds pretty straightforward, doesn’t it? The 2019 regulations were all about improving overtaking, so surely we just look in the historical data and count the overtakes.
Well, no.
In order to pull off an overtake, a driver needs firstly to close on the car in front and then secondly to execute the passing manoeuvre. Imagine if we just count all overtakes and in a particular race we get a ‘middling’ result. That could have been caused by either a large number of ‘overtake opportunities’ in which one car has closed on the car in front, but a poor ‘overtake conversion rate’ ; or the cars all stretching out early on, generating fewer overtaking opportunities, but with a very high overtake conversion rate. How frequently pairs of cars come close enough together to create an overtake opportunity has nothing to do with the 2019 aero regulations and is therefore what statisticians call a ‘confounding factor’. It is only the overtake conversion rate that we are interested in.To see how that can be measured, we must look at the available data.
The available historical data
For many years the FIA has published a great deal of data about each race. The source used for this analysis is the excellent www.ergast.com, a site that allows bulk data to be downloaded in several different forms. The key data here is as shown in the picture: each driver’s position and lap time for every lap of every Grand Prix. This allows us to do some useful analysis but the downside is that anything that happens within a lap is completely invisible, for example A overtakes B but then B re-overtakes A. Thankfully that is sufficiently rare compared to ‘simple’ overtaking that we can take it on the chin.
At the start of each lap we can add up each driver’s total race time and find out how close each pair of cars is. If they are within a given threshold – the DRS trigger time of 1 second was used here – we can declare that as an ‘overtake opportunity’. For each overtake opportunity we can then see if the two cars in question have swapped positions by the end of the lap. If they have, it is counted as an ‘overtake conversion’. To get the conversion rate for the race we simply count the number of opportunities and the number of conversions and divide one by the other.
Getting rid of the bad data
In order to do a fair comparison between seasons, we need to include only genuine racing overtake opportunities and conversions in normal racing conditions. That means throwing away quite a lot of ‘bad’ data, for example:
a. Position changes when the ‘overtaken’ car has lost places by pitting or has just come out of the pits.
b. Position changes where the ‘overtaken’ car has delivered an excessively long lap time compared to the rest of the field, suggesting some kind of in-lap problem like a spin-off.
c. Wet races, which are highly chaotic and usually yield an extreme number of overtakes.
d. Any race at Monaco because its overtaking counts are so much lower than at any other circuit. Including Monaco races would be like rating Olympic sprinters by getting them to do a sack race.
e. The first three laps of any race, due to abnormally high bunching compared to the rest of the race.
f. Any lap where the average (median) laptime is excessively high and doesn’t follow the normal pattern of gradual reduction throughout the race. That normally means either a safety car lap or something like a major pile-up, requiring the other drivers to slow down to navigate round the wreckage.
g. Any lap immediately following one filtered out under the previous condition, because that abnormal bunching continues after the safely car has gone in, often resulting in a short-lived burst of overtaking.
After throwing that lot away, we are almost ready to do some analysis – just one more thing to check first.
Confounding factors
Confounding factors are things we aren’t interested in, but which have an effect on whatever we are trying to measure, like the undulations in the photo distorting the reflection. It should come as no surprise that the biggie is the difference between circuits used in an F1 season. In fact the circuit has a considerably larger effect on overtake conversion rate than pretty much anything else and could swamp the season-to-season variations we are trying to measure if not dealt with in some way.
What we need here is some kind of well-established overtaking-potential factor so we can adjust the conversion rate of each race as if it were run on a ‘standardised’ circuit. Sadly (unless, reader, you know otherwise) such factors don’t exist. Therefore all that can be done is to work out the average conversion rate per circuit across all seasons being analysed and subtract that from each race, to account for the circuit variations. Because that’s only at most 8 data points (2012-2019) per circuit, this is far from ideal but for a multi-season analysis it’s all we have.
However, if we are only comparing pairs of seasons, for example 2019 vs 2018, there is a much better way, described next, so we’ll use the multi-season analysis as an indicator and then firm things up on a pairwise basis.
Analysis and results
Finally – the results.
The central red line in the graph shows the mean overtake conversion rate from 2012 to 2019.
Despite all that dumping of bad data and accounting for circuit variation, we are still left with a lot of apparently random race-to-race variation in each season; this is sport after all, so no surprises there. Because of that ‘noise’, 20 or so races per season is not enough for us to get a good estimate of the true season mean, so we have to produce what are known as confidence intervals, in which we can say there is a 95% chance of the true mean lying. If the confidence intervals between two seasons don’t overlap, we can conclude that the chance of there being a genuine difference between them is more than 95%. Put another way, the difference is ‘statistically significant’.
The story. From the graph you can see that from 2012 to 2016 the conversion rate fluctuated up and down fairly randomly. The big change came in 2017, when the cars and their wheels got wider, and that’s when the designers started to direct dirty air around their cars to reduce drag, and unwittingly cause the problem that the 2019 regulations were intended (partly) to solve. 2017’s overtake conversion rate was significantly worse than 2016, as can be deduced from the gap between the two seasons’ confidence intervals. Let’s now look at 2019. You can see from the graph that the conversion rate was just about back to 2016 levels – a significant improvement over 2017.
As previously mentioned, there is a much more robust method of comparing seasons, using ‘paired’ tests, in which the difference between races at the same circuit are compared across two seasons. Two different statistical tests of this type confirm the significant conversion rate drop in 2017 and the subsequent recovery in 2019.
So it looks like a hats-off to Pat Symonds and his technical team, and if 2021 moves us further in the same direction, the characteristic nature of F1 racing from this perspective should be unlike anything seen in the DRS era so far.
But what about 2018, which looks to be about midway between 2017 and 2019? Well an early regulatory intervention by the new regime in 2018 undid some of the 2017 changes and banned shark fins and T-wings (remember them?). On the basis that anything that helped a car’s aerodynamics in the pre-2019 era probably had a detrimental effect on a closely-following car, it looks like those very early interventions may also have been successful; we just can’t sure from the analysis because of the size of the confidence intervals.
So there we have it. “Did the 2019 F1 aerodynamic regulations achieve what they set out to achieve?”
YES, THEY DID!