First look at the 2016 cars

I was as excited as anyone to see the 2016 cars hit the track in Barcelona. After a winter of boasts from Ferrari, Renault, and Honda about the progress they have made, we finally have a chance to properly assess where they all stand.

Often we are told that “testing times are meaningless”, but it’s important to remember that the teams are using their very limited testing opportunities to gather representative data on their machines. Testing of course has to be analyzed cautiously due to the number of unknowns (different fuel loads, aero specs, etc.), but there is a lot of useful information still present for the careful observer. If you only look at the top times set each day, it’s true that you will learn very little, but dig a little deeper and there are usually rewards.

Long-run data

Thanks to @f1debrief, who posted a total of 41 long runs on their twitter account, I was able to look closely at the data and make some initial assessments of where the teams stand. It’s not a complete listing of all lap-times from the test, but it is the best data-set I’ve found.

Here is a graph of the whole raw data-set, which you can see is a bit of a mess to interpret.

There are drivers on different tyre compounds, and many stints are interspersed with much slower laps (e.g., due to traffic).

I cleaned these data up a bit, first by excluding any laps that were a second or more slower per lap than laps either side of them, since that would suggest they were not representative times. This gave the following graph where you can now easily see the general upward trend in lap times within a stint due to tyre degradation. Gaps in the lines are excluded laps.

Next, I categorized the data by tyre compound used, which you can see in the graphs below.

32 stints on mediums

8 stints on softs

1 stint on supersofts

Again, you can see an upward trend in lap times, with supersofts degrading fastest, followed by softs, then mediums. But notice that there’s a lot of variability, even on the same tyre compound, in (a) the rate of lap-time degradation, and (b) the overall level of lap-times between different stints. If you look, for example, at the soft stints, there is one almost flat line around 92 seconds, which is a stint completed by Kvyat for Red Bull where he seemingly trundled around much slower than the tyres were capable of, but keeping an almost constant time every lap, presumably to assess something other than tyre performance.

Rates of lap-time degradation

Next, I calculated the slope of the lap-times for each stint for each driver, using a linear regression. The actual rate of lap-time degradation, which is the net effect of time loss due to tyre wear and time gain due to fuel burn, will be a nonlinear function, but linear fits performed well enough for stints of this length.

Since most long runs were on medium compound tyres, I’ll first focus on those. Here is a graph showing the average slopes (degradation rates in seconds per lap) for each individual stint on medium tyre compounds.



There is quite a range of performances, from some stints with very little degradation to others losing up to about 0.4 seconds per lap. What I also noticed, looking at the data on a per-team basis, is that the quickest overall stints almost invariably involved the higher rates of degradation, which is perfectly sensible, as those would be the stints where the drivers were actually pushing on the tyres.

Representative stints on mediums

Based on these observations, I selected the stints on medium compound tyres that had a degradation rate of at least 0.2 seconds per lap. This included 8 stints: 1 for Mercedes, 3 for Williams, 1 for Toro Rosso, 1 for Ferrari, 1 for Force India, and 1 for McLaren. These selected stints are graphed below.

Looking at these stints, there’s very little in it between Ferrari and Mercedes. Ferrari have an estimated base time (i.e., on fresh tyres) of 1:25.7 and a degradation rate of 0.38 sec/lap, whereas Mercedes have a base time of 1:26.2 and a degradation rate of 0.30 sec/lap. The difference in degradation rates suggests the Mercedes still has some extra potential that was not being used on that run. Nonetheless, this is quite promising for Ferrari.

What will have the competition seriously concerned is a separate stint from Mercedes, which had a much lower degradation rate (0.13 sec/lap) and a base time of 1:25.7. This stint of Rosberg’s was far and away the most impressive from the first test, and it’s shown below, alongside the higher degradation rate Mercedes and Ferrari laps.


This is a frankly terrifying stint, beginning with the fastest lap-time anyone set on mediums (while carrying at least another 15 laps fuel, which is worth over half a second), followed by consistently fast laps with little sign of lap-time degradation. It is a complete outlier in terms of being both very fast and having low lap-time degradation. Since we don’t know fuel loads, we can hope that we are seeing a final race stint simulation versus earlier stints.

Looking back to the other teams, Williams seemed to set all their three runs under similar conditions, with base times of 1:26.6 to 1:26.8, and degradation rates of 0.21-0.37 sec/lap. Toro Rosso are nearly identical, with a base time of 1:26.7 and degradation rate of 0.29 sec/lap. McLaren are currently sitting a little further back, with a base time of 1:27.4 and degradation rate of 0.27 sec/lap. As for Force India, they have admitted that their focus is 2017 and are timing their wind-tunnel upgrades for that development cycle. For 2016 they have therefore made minor modifications to the 2015 B-spec car, rather than developing a new chassis. This would suggest their headline-grabbing series of short runs on supersofts and softs were probably glory runs, but this long-run stint on mediums is still relatively too slow to be indicative of their true performance. They likely had a much higher fuel load or were running a less competitive set-up.

Representative stints on softs

I did a similar analysis for stints on soft tyres. Degradation rates there ranged from 0-0.67 seconds per lap. Again, the stints with highest degradation were also the quickest ones. This time, I selected only the stints with a degradation rate of at least 0.45 seconds per lap. This included 5 stints: 1 for Williams, 2 for Renault, 1 for McLaren, and 1 for Manor. These selected stints are graphed below.

Neither Mercedes nor Ferrari are present in this analysis, but Williams provide a good benchmark with a base time of 1:25.4 and a degradation rate of 0.67 seconds per lap. In their quickest stint, Renault came promisingly close to this, with a base time of 1:25.5 and a degradation rate of 0.64 seconds per lap. Magnussen was certainly pulling no punches in the first test, taking the car right to the ragged edge. Whether other drivers were pushing as hard, we won’t yet know, so the data are taken at face value.

McLaren were again around a second adrift of Williams, with a base time of 1:26.4 and a degradation rate of 0.56 seconds per lap. Since the Honda engine is still at an intermediate stage of development, this is solid progress from 2015, but they are still some way adrift of the major players. As one would expect, Manor are further back, with a base time of 1:27.0 and a degradation rate of 0.56 seconds per lap. It is, however, impressive to see them within 2 seconds of Williams, given they were scarcely inside 107% last year.

Overall assessment

While it is certainly too early to make any definitive claims about the 2016 season, the long runs from the Barcelona preseason test give us an initial hint of the pecking order. Using the above representative stints, I was able to generate some quick and dirty comparisons between teams. Furthermore, because Williams and McLaren ran stints on both medium and soft compounds, we can create a time equivalence between those compounds. On average, the base time (i.e., when both were fresh) for the soft compound was 0.9 seconds quicker than the medium compound, with a degradation rate 2.2 times greater for the soft compound.

Adjusting for these effects, I compared medium and soft stint times directly. Using this approach, I estimated average lap-times over a 10-lap stint for each team represented above, except Force India. For teams with multiple stints, I took the average estimated time. These times of course have to be taken with many caveats, including the fact that we do not know exact fuel levels.


While it would be extremely premature to conclude that the most severe period of dominance in the sport’s history is over, Ferrari’s new contender looks impressive right out of the box. The concern is that very fast, low-degradation stint from Mercedes, which is the one with an asterisk. We have to hope that stint is not directly comparable to others, or we’re in for a rough season. In any case, Mercedes will very likely ramp up the pace from here, now that reliability has been so impressively established, but we can at least be fairly certain that Ferrari have not completely dropped the ball, which is always a risk with a totally new concept design.

Williams are also staying in touch, but they look to face new threats from Toro Rosso, Renault, and possibly Force India. McLaren are less competitive, but also should not be counted out too soon, given development of the Honda engine will continue into the next test.

What is most impressive to me at this stage is the narrow time range between the best-funded teams and the minnows. Manor look to have made a huge step forward from 2015, and Haas also looked entirely respectable in their first test. If so, this is great news for the competitiveness level in Formula 1. Sauber are impossible to judge until they run their 2016 car, but we could reasonably expect them to be in the mix with McLaren, Manor, and Haas.

The real mystery at this stage is Red Bull, who did not reveal their true pace in any of the posted long runs. As seen in the graph below, their long runs mostly involved maintaining a nearly constant lap-time, while running well below the normal pace. Only one stint from Kvyat showed signs of tyre wear on lap times, and even that was well below normal and ~3 seconds off the pace. These generally look like early race stints with a close to full fuel load and careful tyre management.


If Renault have produced a competitive engine, which Magnussen’s long-run pace would currently seem to suggest, Red Bull would have a very good chance of challenging Ferrari and Mercedes for wins in 2016. Renault’s decision to increase their commitment to Formula 1, and maintain their embattled alliance with Red Bull, would then begin to make a lot more sense. For now, however, Renault are still faced with reliability troubles that Ferrari and Mercedes are not, which could undermine any serious title challenge.

The next test will bring things into sharper focus. For now, hopefully this is an enjoyable quick analysis.


  1. Thanks man! Great article! Definitely provides a different view on it

  2. Malc Harding · · Reply

    Thank you for your hard and detailed work on this, makes very interesting reading

  3. […] First look at the 2016 cars (F1 Metrics) […]

  4. Thanks so much for these insights! Definitely one of the better posts on /r/formula1. Ferrari are looking really promising and I hope that 2016 will be a better season the last.

  5. digitalrurouni · · Reply

    Brilliant analysis. As a very new student of data analysis this article was a revelation to read.

  6. I like what you’ve done here. One thing that could potentially skew the results arrived at here could be the assumption that similar ‘base’ times (your term) imply similar fuel loads. Because ultimately you account for the differences between cars interms of tyre deg and fuel spent. Don’t you think a few KGs of starting fuel load differences between cars could easily amount to some of the differences we’re seeing here? And in a sport were a 0.2-0.3 sec per lap advantage in race pace is enough to be a GP winning margin, what conclusions could we possibly draw from the data? nevertheless, thanks for your sterling work though it was good see some nice graphs and plots to get in to before the start of the next test! Roll on Melbourne!

    1. Thank you, and I agree. As mentioned in the article, we really don’t know about fuel levels. However, fuel-based time differences will tend to be approximately quantized, depending whether a team is running a qualifying simulation (close to zero fuel), a first stint, a second stint, or a last stint simulation. The time difference due to fuel between one stint and another is ~1 second. If cars are doing stints on unequal of fuel, the comparison would be tend to be out by an integer multiple of ~1 seconds. In the case of Force India, I think we can see that.

      Given the fine margins between the other teams, I doubt there’s a multiple of ~1 seconds missing, with the possible exception of Renault, whom I expected to be maybe a second slower than they are. Their position is currently really difficult to be confident about without any useful Red Bull data.

  7. BarryF · · Reply

    This is great, thanks so much.

  8. digitalrurouni · · Reply

    It would be absolutely interesting to read your thoughts and your analyses on the 2nd test given that tomorrow will be their last day. What I am finding interesting is how a lot of cars are not able to switch on the ultrasofts to give them that added advantage over the soft tires. I think it is more related to the track surface being quite abrasive and the temps not being high enough in general.

    1. Thank you! I am busy mulling over the data right now.

  9. Erik S. · · Reply

    Hey, great and interesting as always! I can’t wait for the one after the second test and for the posts beyond that!

    Regarding that, I have two questions:
    1. Do you plan on posting anything on how the top 60 list has been updated with the last few seasons results in account?

    2. I have a vague memory of you mentioning doing something on what-if scenarios, such as what if Peterson had gone to Shadow in 1975, Rindt staying put at Brabham in 1969 or Alesi going to Williams for 1991. Is this still planned?

    1. Thank you! I’m deep into the second analysis and hope to post it shortly.

      1. I’ve been busy working on an upgrade to the model. My hope is to include those upgrades with the next all-time list, although I’m not sure exactly when that will be. It’s a tricky problem!

      2. I do have an article along those lines in draft stage, so it’s hopefully one for the future when I have a bit of downtime — glad to hear there’s interest!

      1. digitalrurouni · ·

        Wow I am looking forward to reading your thoughts!

  10. […] testing finished, I continued my earlier analysis to form a clearer picture of the team hierarchy as we go to Melbourne. This time I have taken […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: