Does Qualifying Matter?
Over the last few years we’ve seen some pretty exciting racing in Formula 1. Short tyres lives, large performance differences between tyre compounds, DRS and KERS have all played their part in making overtaking more frequent and races harder to predict. As a result we’ve seen drivers compromise their qualifying to secure a better chance on race day. Many mid-field teams opted to sit-out Q3 rather than burn tyres setting a time, and it’s not uncommon to see a penalised driver choose to start from the pit lane rather than the grid.
So does a driver’s starting position still strongly affect their race? Or does qualifying just not matter that much any more?
Grid Position vs Finish Position
To answer this question we took a look at cars’ Grid Position (not qualifying result, but the order they started in) compared with their Finish Position (before any post-race penalties were applied) for all races from the 2013 season:
How to read this plot: If a driver always finished where they started you would expect to see a row of squares at 45 degrees from the bottom-left corner to the top-right. Drivers who gain places during the race move down the y-axis (under the 45 degree line), and drivers who loose positions move upwards. I’ve ignored drivers who retired from a race. The colour of each box indicates how many races had a driver start/finish in that position.
So how important was qualifying? That’s essentially measured by the width of the diagonal band running across the plot, and the first impression is that your grid position certainly isn’t a sentence to finish in that position. It’s not uncommon to gain or loose around 3 places, though anything much more is unlikely.
The least amount of movement is found towards the front and back of the field. At the sharp end, the top 5 or 6 qualifiers tended to hog the top 5 or 6 places, even if they did swap around amongst each other. At the back is a similar situation, where Marussia and Caterham found it difficult to break into the mid-field, and instead diced amongst themselves. In the middle there’s more variability and uncertainty, with large gains and losses in position possible. Here, there’re more strategies to keep track of, more traffic to get caught up in, and more excited and less experienced drivers to battle with.
The leading 5 or 6 cars tend to monopolise the top 5 or 6 positions, but they swap around with each other in doing so. With a few exceptions, if you wanted to finish on the podium you needed to start in the top 10, and if you wanted to win you needed to start in the top 3 (although that statistic is strongly affected by Vettel, see below).
In answer to our question then: a bad qualifying wasn’t the end of your race, and a good driver in the right car should have been capable of making some progress. However, the scale of that progress is limited, and achieving the result you wanted was probably predicated on getting a good time in qualifying. So, qualifying still matters.
Some other observations
The first feature you’ll notice is the bright box for starting on pole and finishing in first. This happened 10 times in 19 races during 2013, and 8 of them were a result of Sebastian Vettel’s domination over the rest of the field. Regardless of whether you’re a Vettel fan or not, I think we all hope that this feature won’t be repeated in 2014.
You might also have noticed that there is a bias. This can be seen more clearly if we draw on the 45 degree “you start where you finish” line:
While at the sharp end of the grid, the distribution is approximately symmetrical about the yellow line, the further down the grid you go the more likely you are to improve your position during the race. This is primarily because of retirements during the race. As cars drop out you move forward, and the more cars there are in front of you the more likely this is to happen. Unfortunately for the teams at the back, reliability was pretty good last season (something we’re less likely to see this year!) and, from a rough visual inspection, the shift forward seems only to be worth about 3 places a race. By the time you get to around 10th place on the grid, the bias is hardly noticeable.
One feature that isn’t present is a discontinuity between cars starting in 10th and 11th. While cars which are a single place apart should be fairly evenly matched on pace, a car starting in 11th has much greater strategic freedom: not only do they have an extra set of tyres which weren’t used up in Q3, but they can also choose which compound to begin the race on. One might expect then that they have a greater chance of moving forward during the race than their 10th place counterpart. However that doesn’t appear to be the case. This may be because of grid penalties from cars qualifying in the top 10, meaning the car which qualified in 11th didn’t always start there – but this would require a more dedicated analysis to tease out.
One of the longer term ambitions surrounding this blog is to use the observations to develop a simulation which is able to match, and then hopefully predict the outcome of Grand Prix. Since this plot turned out to be relatively simple, I made a simple attempt at a race simulation and comparied the outcome.
The simple simulation involved races of 60 laps. Each lap, neighbouring drivers had a 5% chance of swapping positions (overtaking) and each driver had a 0.3% chance of retiring (I picked those numbers from my imagination). For a 19 race season here’s what that gave:
While it doesn’t look quite the same as the real thing, it correctly reproduces the main features: the diagonal band with width of around 3 places, and the gradual bias due to retirements. Probably as expected. What this model doesn’t reproduce is the greater uncertainty in the mid-field, or the outliers from fast cars having bad qualifying and coming forward, or vice versa. And it doesn’t reproduce Sebastian Vettel’s dominance.
Perhaps including pit stops and tyre strategies, or the potential for non-fatal damage will help explain the mid-field variability. While the outliers may require accounting for the different abilities of the teams and drivers.
Regardless, this appears to be an encouraging first step given the simulation’s simplicity.