Le Mans data analysis, part 2: Picking apart driver averages

The next part of The Racing Line's analysis from the 92nd edition of the 24 Hours of Le Mans picks apart the driver averages in Hypercar to find some trends.

The methodology here is the same as the previous part: identifying's driver's top 20% fastest laps and then averaging them out.

In theory this should eliminate any laps done under yellow (either local yellows, slow zones, or full course yellows), and the safety car.

However, as you can see in the graph below... this has created some outliers. This was caused by the four-hour safety car in the race, from 3:44 to 8:10 in the morning. Drivers kept circulating behind the safety car and these laps were obviously recorded every time a car crossed the timing loop at the start/finish line.

If drivers did a lot of running behind the safety car then, or elsewhere in the race, these laps may have crept into the 20% average.

Either way, the full chart is below. We're going to break this down into sections as it's a big graph.

And the fastest driver was...

At first glance, this is a surprising one. However, the data doesn't lie: Dries Vanthoor, in the #15 BMW, was the quickest driver when taking into account top 20% laps, averaged out.

The Belgian driver set a 3:30.281 to record the quickest average.

This is somewhat surprising because the #15 BMW was way down and out of the fight almost all through the race. However, given the car retired after roughly 7 hours, at 9:30 or so local time... the lack of data means this sort of makes sense.

Indeed, Vanthoor only completed 30 laps in the car. 20% of that is a grand total of six laps. So this probably can't be taken as truly representative. It does signal BMW's good pace and the strength of their drivers though.

From cars that completed the race, Brendon Hartley, onboard the #8 Toyota, was the quickest average driver, with a 3:30.372. His teammate in the sister #7 car was next up, Jose Maria Lopez, just a few hundredths of a second off. They were followed by the fastest average Ferrari driver, Alessandro Pier Guidi in the #51 Ferrari, 3:30.540.

On it's own, this doesn't tell us that much. But it does confirm that the Toyotas were, in general, quicker than the Ferraris.

Clearly, some of the times are outliers, possible due to running in wet weather, or maybe due to the safety car periods through the race. Indeed, Nielsen did almost three entire stints under the safety car through the race, one late on Saturday evening and the other two through the four hour safety car period from 3:44am until just after 8am on Sunday morning.

Otherwise, consistency is good, especially at Ferrari, barring Kubica and Nielsen as outliers. If we look at best laps, we can see the times are consistent, as you'd expect for single lap pace.

A few bullet point notes from this chart, which shows the fastest Toyota, Porsche, Cadillac and Ferrari drivers over a single lap:

• The top 3 fastest drivers over a single lap in the race were from Toyota drivers.
• Alex Palou was the quickest non-Toyota and, by that fact, the quickest Cadillac driver. He was also the second-quickest Cadillac driver in terms of top 20% average.
• Ryo Hirakawa was the slowest over the single lap from these drivers. He was also the slowest Toyota driver when looking at the top 20% average, but this was marred by him doing lots of running under safety car and rain.
• Alessandro Pier Guidi was the quickest Ferrari driver over a single lap and in top 20% average. Yifei Ye was similar in the #83 satellite AF Corse Ferrari entry.
• Estre was also similarly the fastest Porsche driver in both metrics. Fred Makowiecki was the slowest, but Andre Lotterer was not much faster.

Pulling apart the skewed data

The skewed data, when looking at average lap times, is obviously a little bit problematic because it doesn't give a complete picture of the averages.

It's not as easy as removing the laps done under safety car, either. Because there were so many laps completed either under yellow flag conditions, with slow zones in place, or under safety car, removing them gives a very small sample set to work with.

And that's not even counting laps done on a fully green track but in wet or changeable conditions.

Even so, for the worst-affected driver, #87 Toyota's Ryo Hirakawa, I've attempted to align it so we get a more complete picture.

So, Hirakawa did 77 laps in the race, over two driver stints and a total of seven fuel stints. After his outlap on the first stint, he set four consecutive laps under green flag that were relatively quick, the slowest being a 3:33.267.

After this he set a 3:36.611. While the timing data says the next set of laps were under green, because it doesn't record slow zones, only FCYs or safety cars, they were all slower, in the four or five minutes range.

After this, apart from one lap, a 3:36.223, his 17th lap on his first of the two stints he was in the car, the fastest lap he set was a 3:49.705.

So, the sample of truly representative data is four consecutive laps at the start of his first stint, the average of which was 3:32.699. Obviously his fastest lap, a 3:31.917, was also set in this sequence, the final lap before that 3:36.611.

Examining Porsche driver performance

So, as already mentioned, Estre was probably the quickest Porsche driver overall, with the Frenchman the quickest in both aspects.

Phil Hanson, in the #38 JOTA Porsche, was the quickest non-factory Porsche driver, again in both aspects. In fact, he was second quickest behind Estre. Callum Ilott, in the sister #12 JOTA, was also quick, as was his teammate Will Stevens.

Julien Andlauer, who was so impressive at Spa, was also quick, with a very good average lap time.

In terms of factory drivers, Felipe Nasr was among the slowest. Lotterer had a slower fastest lap, but his average was quicker, unusually. Makowiecki and Jaminet were also unusually slow according to these metrics, but again it's possible they did a lot of running in the on/off rain or under local yellow, slow zone, or FCY.

It's interesting to note how many of the Porsche customer drivers were quicker than the Porsche factory drivers. We'll say no more on the subject... for now...

Schumacher fastest again at Alpine

Mick Schumacher was quickest amongst the Alpine drivers, in terms of average lap times at Qatar and Imola. While I haven't had time to look at the data from Spa-Franchorchamps – life getting in the way — he was quickest again, in terms of both averages we're measuring, at Le Mans.

The German was over three tenths quicker on average compared to the next quickest Alpine driver, Charles Milesi, two and half tenths quicker in terms of best lap than Paul-Loup Chatin. Both drivers were in the other car; in his own car, he was easily faster than his teammates Nicolas Lapierre and Matthieu Vaxiviere.

Something that looks like an anomaly here, but actually isn't, is Ferdinand Habsburg's laps. His average is exactly the same as his best lap. Huh?

I've double checked this and it seems the data is correct. The Austrian only managed five laps in the #35 Alpine before the engine gave up and he stopped at Arnage. His last lap was his fastest, but it seems if you add up the lap times of those five laps, it's exactly the same, to the thousandth of a sector, to his last and best lap.

I'm no statistician but that seems incredible. I really hope I didn't get my sums wrong...

Update: the Habsburg thing is interesting, looking at the data and my formulas in more detail.

He did five laps before the car gave up and he retired. Those laps were 5:07.499, 3:34.287, 3:33.728, 3:32.772, and 3:32.320. The car gave up on his sixth lap, meaning it wasn't recorded.

My formula to create averages looks at the top 20% of lap times a car or driver set. Because Habsburg only set five lap times in the race, his top 20% was just one lap, the last one and his fastest, 3:32.320.

So the sums and the formulas were correct... but the threshold wasn't enough. If you widen the threshold to the top 80% of laps, so all four laps apart from his outlap, you get an average of 3:33.277.

A final note: as the Alpines both retired early on its impossible to say if Schumacher would have continued to be the fastest Alpine driver. The data only shows us if he was when the cars were in the race (obviously).

tl;dr

• The rain, amount of safety cars, and yellows/slow zones skewed the data a little bit. I may try and fix this to have another look later... but that's for another time.
• When looking at 20% averages, Dries Vanthoor in the #15 BMW was quickest of the Hypercar drivers.
• Of cars that finished the race, Toyota drivers were the top three quickest.
• #7 Toyota's Jose Maria Lopez set the fastest lap of the race, towards the end of the 24 hours.
• Kevin Estre was the fastest Porsche driver when looking at both top 20% average laps and fastest laps.
• Mick Schumacher was the fastest Alpine driver when measuring by the same metrics, but it's impossible to say if he would have been had the cars continued.