F1 Data Junkie – Driver DNA

Although I missed the live race for the second time in a row, and didn’t get a chance to play with the data as quickly as I would have liked to, I did spend some of my time away wondering how to plot all the telemetry data for a driver captured during a race in a single graphic.

The single lap view, like this one from one of Button’s laps at the 2010 Malaysian Grand Prix:

is all very well, but if we overlay traces from each lap onto the distance labeled x-axis, the charts just become messy to read.

So how about this instead. On the x-axis, we have the distance traveled round the track per lap. The drivers are pretty consistent in the lines they take, so the overall distance is pretty consistent. If we have a 4km track, and a chart that’s 400 pixels wide, each pixel corresponds to 10m resolution of track distance. For the y-axis, we use the lap number. And to plot the actual value of a telemetry measurement, let’s use colour. Put these together, and we come up with some driver DNA charts – the ones below are form Hamilton:

Hamilton's Malaysian DNA

So how do you read these? Each strip is a different measure. The colour intensity increases with increasing value up to the maximum recorded value. Within each strip, time flows down the strip.

The top, blue strip shows the gear (1 to 7); the green strip shows the throttle pedal depression (0-100%), and the red strip shows the brake (0-100%). The light blue strip is a composite of the previous three strips. The whiter the pixel, the closer it is to 100% throttle in 7th gear with no braking.

The bottom two traces show the longitudinal and lateral g-force respectively. For the longitudinal trace, red shows braking – being forced into the steering wheel; green shows acceleration – being forced back into your seat. You’ll see the greatest g-force under braking occurs when the brakes are slapped full on… (the red bits in the third and fifth traces line up). For the latitudinal g-force, the red shows the driving being flung to the left (i.e. right hand corner), the green shows them being pushed out to the right.

I’m slowly pulling enough tools together to be able to start telling some stories… so stay tuned ;-)

Author: Tony Hirst

I'm a Senior Lecturer at The Open University, with an interest in #opendata policy and practice, as well as general web tinkering...

7 thoughts on “F1 Data Junkie – Driver DNA”

  1. Interesting. I wonder if you can come up with the perfect DNA for a course, and then match driver to how close they are, and whether their final race position correlates well (I guess that’d indicate how much other randomness is in the race – if the DNA predicts the race well, then it was probably quite processional…)

    You could imagine tracks with lots of straights and sweeping curves (Monza, say) having a very different optimal DNA from tracks with lots of twists and turns (Monoco, say).

  2. Interesting. I like the map-based graphics more, though, just because they’re prettier to glance at, as someone with no real knowledge of this field. I wonder if you could draw colored lines around the track, using the same brightness-as-value thing. It would probably take lots of tweaking to get it to match the right path, though. And of course which ever lap got mapped to the outside of a corner would be stretched the most and the inside squished up, and it might end up having to be so wide as to be useless.

    But of course this really brings out the links between the different datasets, with breaks lining up with the braking in almost every graph.

    I was just reading a blog post by (one of?) mclaren.com’s web developer(s) about the system behind this, and wandered onto your site by googling to see if anyone was making use of the raw data. Then I finished reading his blog post and saw he linked to you directly!

    You might be interested in the series of articles at and around http://kenneth.kufluk.com/blog/2010/04/building-mclaren-com-%E2%80%93-part-3-reading-telemetry/ .

    1. Rather than doing lines between points, I was thinking of doing a heat map around the track, or at least on the approach to and run off from the corners. With a bit of luck, I’ll get an hour or two to hack a demo of that together over the next few days…

      It also struck me that from driver timing differences (e.g. as charted on http://www.f1fanatic.co.uk/2010/04/08/funky-new-interactive-race-charts/ ) and the time based samples I get for HAM and BUT from Mclaren, it should be possible to guesstimate the positions of *all* the cars on the track once per lap? I suspect it would take a bit of careful thinking about how to do it the first time, but then it should just work…;-) Not sure how useful this would be, but it’s an interesting puzzle about how to actually implement it in a painless way…

      Re the blog articles, I hadn’t seen that – so thanks for the pointer. Btw, any further comments about the F1 visulaisations I’ve been posting, and any questions you have about the races that you think might be answered or at least illustrated by visualisations, I’d love to hear them. Once I’ve got a couple more things in my toolkit sorted, I hope to start trying to use the visualiastions to teach myself (through a series of public blog posts) a little bit more about how F1 cars get driven round the racetrack, and also feed into a revised version of my old and aborted F1 Race Day Strategist spreadsheet ( https://ouseful.wordpress.com/2009/06/06/mulling-over-what-to-do-next-on-the-f1-race-day-strategist/ ).

  3. Great work and I like the principle of this vis, but the resolution in the data limits how much I find I can get out of it.

    Eg, F1 is all about braking points, speed carried through corners, traction out of corners etc, and very little about the long accelaration in the straights which take up most of the horizontal scale, and the corners happen so quickly in this data that it doesn’t show much detail to see the lap on lap differences.

    Also difficult to compare late laps (low fuel) with early laps with this layout.

    However, my problems seeing details could equally be due at least in part to just how phenomally consistent drivers are lap after lap. :)

    Would be useful to see more easily in the vis the tire changes (I’m assuming the darker like horizontal lap would correspond, getting tires up to heat etc) and which sectors were setting purple or green times (perhaps another use for colour rather than just intensity if you can pull that data from other source perhaps?)

    it looks facinating though and comparing 2 drivers in the same team is probably going to be more revealing of the drivers styles than would be apparent across teams.

    Now, if this could just be overlaid on screen in real time with dynamic comparisons and across practice/qualification/race pace… ;)

    1. Hi Nick –

      I quite agree about the resolution of the data not really being fine enough to properly see what’s going on, but whilst doodling with a couple of ideas for comparing HAM and BUT, one of the traces identified driver differences on part of the Malaysia lap (I’ll post more on that in a day or two…).

      On the to do list is a pairwise lap comparison view, but I’m not sure of the best way (any way!) of doing that yet…

      Also on the to do list is to supplement the Mclaren telemetry data with published timing stats. As far as tyres go, is it possible to get a record of which tyres go on to a car in which lap from anywhere? (Related to this, is there any other data out there that I can use to supplement the telemetry data?)

Comments are closed.

%d bloggers like this: