Malaysian Grand Prix: Story from the Data

Posted on March 26, 2012

8


Races like the Malaysian Grand Prix are why I got interested in motor racing in the first place. Varying conditions, different cars fast at different stages of the race, great overtaking moves, visible driver skill and surprise standout performances. It makes the job of analysis much more difficult, but then I only ever started playing with the numbers because I love the racing.

One other very pleasing thing at Sepang for me was the performance of Sergio Perez. The Mexican has long been a bit of a data analysis hero on this site, with the underlying pace analysis consistently pointing to his being a potential superstar, although this has not been backed up with enough hard results. Although I did not expect him to be on the podium for some time to come, a big result has been coming, and I would expect him to back it up with some more great performances whatever the colour of the car he may be driving…

Since I developed the model and started analysing the races (Monza 2011) , we have not had a substantially wet race, so there is some playing to do to get the most understanding from the laptime data, and the fits are unlikely to be as neat as usual due to the constant evolution of the track surface – which is extremely non-linear, and not modelled. So we will have to see how much I can extract from the data and whether existing models can be tweaked to do the job.

What I have done is to split the race up into three segments – up to the red flag, the long middle period of the race on intermediate tyres, and the final act on slicks. Each has quite different characteristice, and the last segment is the easiest to analyse. But we’ll start at the beginning. The race history chart for the first few laps (with some notes) looks like this (and is it quite messy).

What can we see? Well, firstly, the track was (obviously) getting wetter, and the laptimes were becoming much slower. We can clearly see the rate of decrease of the pace of Hamilton (on inters – top black line with crosses) against Perez (on wets – orange line with crosses). Pace is represented by the gradient of the line, and we can see that the pace of Hamilton reduces very quickly (multiple seconds per lap) whereas the pace of Perez on the wet tyres is much more stable as the track gets wetter. Interestingly, the gain from Perez is not as great as might have been anticipated, being less than 10 seconds on Hamilton, as the real benefit of the wet tyres is only really seen from lap 4 (when Button stopped).

However, it is clear that the early stoppers benefitted – the position of Massa before and after the stops is compared (up 2 places), and Alonso similarly jumped Vettel – something which turned out to be useful later. The real gainers from the red flag, though, were those who stayed on their original tyres – Vergne (inters) and the HRT cars (wets). With no time lost for a pitstop, Karthikeyan (shown on chart) went from being at the back, to being in the middle of the pack, as Button was to find out to his cost.

In terms of the relative pace of the cars, there is not that much to learn from the first stint. So let’s move on to the second, and pivotal segment of the race. What I have done here to model the pace of the cars is to use the normal tyre degradation model for the first part of the stint (10 laps or so) where the conditions are fairly stable, and to use the nonlinear phase 2 tyre model (in reverse so the track gets faster) to model the improvement in the track conditions. This gives two pieces of information – the pace in the wet, and the relative improvement in pace (due to changing conditions and tyre state) as the track becomes drier. As long as the fits are sufficiently good, this should tell us something about the car performance, and perhaps something about the drivers too. First up, I have a plot of the first five drivers (in the stint – not final order) and the fits I have managed to get for this stint.

The fits are much better than I had expected, but there are a few variables involved. The initial curve is modelled with the degradation model (linear) for a number of laps depending on the car. At this point the pace picks up, and a quadratic term is added for the pace pickup. This works really nicely and we can see that the initial pace of Alonso is significantly quicker than the others, but but he is the last to pick up the pace, and also picks up his pace by the smallest amount. Such was his initial pace (0.5s up on Perez, 1s up on Hamilton/Red Bull), though, that he only really loses significant time to the Sauber, who is improving his (fuel/tyre age corrected) laptimes by 0.45s per lap, when Alonso is gaining at 0.33s per lap. After 10 laps, Perez has overturned the 0.5s per lap time loss, and by the end of the stint is catching Alonso by 1s per lap. Stunning.

Fitting Hamilton’s curve required his being a full second per lap slower than Alonso, but he was earlier in picking up the pace as the track dried, and his laptime improvement was better than the Ferrari, but way down on the Sauber. This left him with a stabilised gap, but at no stage was he really making significant inroads. The Red Bull curves are essentially the same as the McLaren, but they are held up by Rosberg at the beginning of the stint. Rosberg’s tyre struggles can be seen clearly, and he stops for new tyres at around the time the track improvement can be seen clearly. Before the stop, he was 2s per lap slower than Alonso, and after, he is much faster (which shows that the degradation of the tyres is of the order of 2s within about 10 laps, but it is offset by the improvement in track grip), but the rate of increase of pace is much slower so he continues to fall away. He was better off stopping, but only because it was clear that he was in real tyre trouble. Curiously, Button is much the same – he is 0.2s slower than Hamilton after his nose change, which seems OK, but as he wasn’t picking up the pace when Hamilton (and Maldonado in front of him) was, he pitted. His pace became much faster, but again the pick up as the track dried was much less significant on the newer inters – once the track got close to dry, the more worn (closer to slick?) inters were significantly faster than the newer tyres. Button would almost certainly have been better off not stopping, but realistically he only had the pace to be in the hunt for a point or two – mainly due to the field being so closely matched.

Of the others, there are a few interesting notes. The first is the performance of Kobayashi. He shows the same improvement per lap as Perez, but he started off the stint a whole 3.5s per lap slower. It makes the Mexican’s performance all the more remarkable.ย  The same goes for Alonso – Massa is a full 2s slower than the Spaniard, and he kills his tyres too. He is also very slow after his stop – it’s difficult to believe that his drive is not under threat. Also, worth mentioning is the performance of the Mercedes. In contrast to the Sauber and the Williams which gained significantly as the track dried, the Mercedes clearly struggled to keep their intermediate tyres alive. At the end of the second segment of the race, they were the slowest cars in the midfield. These are shown in the chart below.

This second segment of the race could have run significantly longer had Ricciardo not decided to chance a stop for slicks. His laptimes were clearly fast, so everyone else stopped too – within a few laps anyway. The race history chart for the final segment of the race is below. Modelling is again difficult due to the improvement in the track conditions, but there is a normal linear behaviour in the last ten laps (or so) from which the underlying pace can be picked out.

If you watched the race, the pace of Perez was clear, but he wasn’t the fastest. That was Kimi Raikkonen, and his fastest race lap was an accurate reflection of his pace. The other standout is the pace of Bruno Senna in the Williams. When it comes to race pace, Sauber, Williams and Lotus were good in Australia, and they are good here. Of these teams, it is the Lotus cars which are starting closest to the front. There must a big result for the black cars on the way. It would be nice to see Williams on the podium as well. From this evidence, it’s not out of the question, especially when we look at the unconvincing pace of the McLaren and Red Bull cars in this final stint.

So of the cars with clear air, and therefore representative pace, we have the following pecking order (using Webber as reference):

Raikkonen -0.5s

Webber/Perez +0.0s

Senna +0.1s

Alonso +0.4s

Hamilton/Schumacher +0.5s

Di Resta/Vergne +0.7s

Massa +1.0s

It makes for fascinating reading. Of the ‘fast’ cars, it is only Webber who gets his act together in the last few laps, although he was not so quick earlier in the stint. Hamilton appears to have settled for third place, or doesn’t have the pace – maybe a combination of the two. Alonso maintains a pace slightly faster than Hamilton, but is slower than the Sauber and the Williams. The standout car, though, and by a clear half-second per lap is Raikkonen’s Lotus. Mercedes show reasonable pace, as do Force India and Toro Rosso. And bringing up the rear of the midfield for the second consecutive race? Massa. At what point did he find out that his team mate won the race?

It is easy to imagine that the big boys lost this race, and that they shouldn’t have allowed two drivers who scraped into Q3 to run away with it. But they didn’t throw it away with errors. Yes, Button and Vettel both hit Karthikeyan, and Hamilton lost time in the pits twice, but that is not the reason they lost the race. The reason, as admittedly it usually is, is that the cars in the first two places were simply faster – especially in the long stint on intermediates. Add in Raikkonen in the final 10 laps, and we have three unexpected cars showing the best pace. With many competitive cars, and different cars fast at different stages of the race, Formula One 2012 is looking very promising indeed.

Advertisements