As the second (and final) pre-season test at Barcelona is now underway, this will the last post on the first Barcelona test. Two sets of long runs were particularly noted in the comments, demonstrating how difficult it is to assess the pace of the cars from simply the laptime data. These data belong to Massa and to Rosberg. The idea of this post is to present the graphs of the data in such a way to show what I was thinking (rightly or wrongly) in my interpretations of the data, and to make it a bit clearer how I have tried to understand the data.

So firstly, Massa. The question here is whether he refuelled after his first run. What I have done is plotted his three runs on the equivalent of a race history chart (in red), plotted the intelligentF1 model curve fits (should match the gradients of the data – the time lost/gained in fictional pitstops is not relevant here) and as a check, I have plotted the second stint (minus slow first timed lap) over the first stint in blue – these should be close if the Ferrari is refuelled to it’s original fuel load.

So what we have is a first stint which has two speeds – this is not a typical trace and this step change is not tyre related. A reasonably consistent fast pace for the first few laps, and then a slower, but reasonably consistent pace for the remainder. Rightly, or wrongly, I chose to fit the slower pace as usually in the races, this is more indicative of the true pace over a stint. The fits for the second and third stint are pretty good – the trace at the end of the second stint is more indicative of the tyres going off more quickly as the trace gradually bends away from the fit. The curvature of the line gives the difference between the gain in pace due to fuel burn and the loss in pace due to tyre age. Generally, the stints are getting slower, so the tyres are losing pace faster than the gain from the effect of fuel load reduction.

The fit in the graph for the first stint requires a pace 0.6s faster than for the second and third stints (identical) using the fuel effect of 0.11s per lap which I derived from the 2011 race by matching stint paces for the same car on the same tyre type/age for different stints and then averaging over the cars. From the opening few laps, it could be argued that the fuel loads are the same in stints 1 and 2 as the blue and red curves match. However, it could also be argued that the second stint is faster – not by enough for it to be purely fuel effect, but by enough to argue that it could be something else like a different tyre compound. The truth? We will probably never know unless they decide to tell us.

The complexity in Rosberg’s trace, however, is somewhat different. Rosberg did four stints, and a comment on the last post has it that the Autosport commentary has all four stints on hard tyres. The timing data I have also suggests that Rosberg did not go into the garage between the first and second stints. The graph of Rosberg’s race is below.

Again the pace is given by the gradients. The dashed line is the mode fit, and the fit for the first two stints is at a pace 1.2s (not 0.6s as I incorrectly stated in the previous post) slower than for the last two stints. It is immediately obvious that the second stint does not follow a typical trace. It looks like two separate stints – a fast pace to start, and a slow pace to finish – neither of which are consistent with the first stint pace. On average, the early pace could have taken more from the tyres resulting in a lower pace at the end. This pace change is usually seen in fuel saving modes in races – the signature on the traces is quite clear, and it is possible that the second half of this stint is a test of a fuel-saving mode. Given that the pace difference between stints 2 and 3 is 1.2s (fuel corrected) and the match from stint 3 to stint 4 is very good, the suggestion that Rosberg refuelled between stints 2 and 3 is probably correct – even though the fuel load at the start of stint 3 would be slightly lower than the load at the start of stint 2 by my reckoning.

I hope that this sheds some more light onto where I get the numbers from, and how I have reached the theories on what the teams have been doing. Of course, all this is speculation, but it I hope that it can give some insight into what is happening in the tests. I’m glad there’s no refuelling in races – it makes it much easier!

*In Depth Data Analysis, Underlying Pace Analysis*

James Porter

March 1, 2012

Very good explanation, keep up these analysis.

I’m still unsure on the Massa analysis, though this is probably the most likely answer. I think you’ve pretty much hit the nail on the head with Rosberg. I don’t think we can deduce anything else from the data that is available.

One question: Could fuel saving in Rosberg’s stint 2 have allowed for a relatively heavy weight for the start of stint 3 making it look like re-fuelling had taken place? I don’t know what level of fuel they manage to save.

intelligentf1

March 2, 2012

It’s hard to tell what level they save, but they only lap slightly slower (2% is about 1.8s), so it won’t be much. From the races I did last year, I have kept the same fit for fuel effect even when it is clear that fuel saving is going on, and I don’t get anomalous results. So I would be surprised if he could save enough for us to be able to see it with this level of modelling or the amount of data we have.

Hello

March 1, 2012

Good plots, I would’t try to look more into the 0.6sec delta that you are seeing, mainly because if for a given reason MAS has used DRS for a few laps he will appear 0.6 faster and we won’t at first glance know why. The best thing to do in this case is to use what you know for certain, that he didn’t refuel on the last two stints…

Can’t you make your model differentiate in beween tyre deg and fuel compenastion? I believe JA posted lap fuel consunsumption per circuit, that should remain constant one every track (and probably safe to assume for all competitors). Once you fuel compensate the laps you can fit for deg only. Once you have both you have the whole picture. For testing it won’t help you much, but I think that for the races you will get a much more accurate fit.

What do you think of today’s running?

intelligentf1

March 2, 2012

With the model I can differentiate between tyre deg and fuel consumption as soon as you get multiple stints on the same tyre. For the Barcelona race last year I get the fuel figures, and then I can calculate the fuel consumption. There is a ‘live’ version (prototype) which will calculate the degradation (each car) based on a fuel effect (calculated average) as the live timing data comes in – it’s Javascript/Flash based so I can’t stick it on this site. I don’t know if you’ve looked back at the races I did last year – I started with Monza and I’ve posted on each of the races since.

Interestingly, I understand that the fuel consumption data from JA is an average across data from teams. What I have found is that the figures I get can be up to 30% different from his, but the average value I get across cars pretty much fits everything OK. I would estimate (very tentatively) that I’m +/- 10%. There are odd anomalies, but generally it’s good. I use linear fits for both tyres and fuel (equivalent to sticking a straight line through the laptimes) as I don’t think the amount of data justifies more.