As we approach the last part of a memorable 2020 Formula 1 season, we’re excited to introduce our final F1 Insight powered by AWS of the year. This year, we have focused on many different aspects of performance and competitive analysis. We’ve looked at the detailed aspects of car performance and how these are broken out into the most important elements. Our final insight of 2020 concentrates in a similar way on the driver’s performance. Having talked many times about how Formula 1 is the perfect blend of human and machine, we want to be able to understand the important elements and factors that make up a driver’s overall advantage.
The Driver Season Performance graphic will concentrate on that very thing: looking at the most important elements that make up a driver’s performance over the season such as qualifying pace, race pace and how each driver is able to manage their tyres to give three really important example parameters. When Formula 1 teams are choosing race drivers, they concentrate on all of the different aspects that we have encapsulated into this new graphic. It tends to give an overall picture of how skilled the driver is in each of these different crucial areas. The summation of all of these elements gives us a complete driver performance analysis and we are then able to compare, holistically, which of the drivers consistently out-perform the others - not just on an overall level, but in each of the individual areas as well.
With this new F1 Insight powered by AWS, we have broken down the main skill sets for a Formula 1 driver and presented them as individual elements that will give us a picture of the complete driver. Each of these seven individual elements is crucial for a Grand Prix driver. They need to be incredibly quick on a Saturday afternoon in qualification, when it is about extracting 100% from the machinery underneath them, but also have all the racing elements affected too. This begins at the race start - which is an art itself - and goes on to look at how the drivers perform on the first lap of the race, i.e. how many cars they are able to overtake or how many times they are overtaken, how they perform in the pitstop situation and how they are able to manage their race pace and tyres - all important elements for the view of the complete driver.
With the Driver Season Performance approach, we are trying to use all the data available between the car telemetry and timing data. This includes the estimated delta pace between cars, car performance data from the race start, delta pace between tyre compound, etc.
We have defined seven different metrics critical to driver performance to give us an overall rating, each scored on a value of 0-10. These are calculated at each race event. Averaging these results race after race give us the ability to increase the number of data points and therefore a better approximation.
We see from TV that some drivers are quicker than others on the start because they are able to better control the wheelspin, and that gives a perception of which driver looks better than others. To evaluate this ability, the 0-100 kph time is evaluated, which represents the very first part of the race start as 100 kph is a speed that is always reached before the first corner.
In order to identify with precision, the starting moment, the longitudinal acceleration signal from the on-board accelerometer is used. There is a bit of a delay before we see the speed moving from 0, but the accelerometer can identify that moment.
The final point is the exact point the driver’s speed reaches 100 kph.
This is another important skill in the initial phase of the race. During the first lap, some drivers will gain positions; others will lose positions in a zero-sum game. The metric will take into account the net result, which is the difference between the number of positions gained and the number of positions lost.
In this process, all the overtakes are analyzed in order to count only the manuevers that have not been affected by crash, mechanical failure and so on, which means only the genuine overtakes are taken into account.
In order to give a score to this metric, it has been decided to assign a specific score to the drivers who were able to maintain their position and add points if a driver was able to improve his position or remove points if he lost positions. Since some tracks allow more overtakes than others because of their design, this has also been taken into account, giving more or less points for the same number of positions gained or lost.
Overtaking during the race
Who is the best driver when it comes to overtaking? Similar to other metrics, in this case, we (the fans or viewers) all have a feeling and/or an idea on the best drivers, but this judgment is usually based on some specific event when a special overtaking manuever has been completed. Unfortunately, this doesn’t take into account all the other overtakes and, even more importantly, is not an objective view we can use. In fact, there may be drivers who are not doing spectacular manuevers but are very effective at overtaking other cars during the race on a consistent basis.
Even if we take into account the number of overtakes during a race event, we are not 100% fair, as drivers with a very good car who for some reason are in the back of the grid, might be able to overtake a larger amount of cars without serious effort. The way we take this into account, is to give a different weight to each overtake manuever using the lap time in the previous 3 laps of the two cars involved. For example, if a driver is able to overtake another one but he was significantly quicker in the previous 3 laps, this overtake is counted less than 1 because it was easy; instead, if the two drivers were very close in terms of lap time, this overtake is counted more than 1 because the overtaking driver didn’t have a substantial gain from his car to count on.
In doing this, we have two metrics; one is the pure number of cars overtaken during the race, the other is the number of cars overtaken during the race but weighted accordingly to the car pace. The higher the difference between the weighted number of overtakes and the pure number of overtakes, the higher the score awarded to the driver. In this way, we can take into account both the number of cars overtaken and the associated difficulties.
This metric is designed to answer the question “who is the quickest driver in qualifying”? To do this, we need to try to put all the drivers in the same “conditions,” which means the same car, same track condition, tyre compound etc.
The first step of the process is to identify which lap times to take into consideration. In order to have as many drivers as possible in the same conditions before applying any correction, we use Q3 lap times for the driver who passed in Q3, Q2 for the driver who reached Q2 but missed Q3 and Q1 for the rest. Among these lap times, the best is used. In case of a driver mistake or inability to perform a proper lap in that session, the lap time evaluated is taken from the previous session. For example, if a driver missed both attempts in Q3, then the best Q2 time is used.
In order to compare the drivers in the same conditions, we have to apply a lap time correction for the drivers who run in Q1 and Q2 in order to make them comparable to Q3 drivers. To do that, the track evolution across the session is calculated and the effect is added or subtracted from Q1 and Q2 drivers.
Moreover, not all the drivers may have completed the lap time with the same tyre compound. For example, a driver who passed in Q3 with Medium compound and wasn’t able to run in Q3, has a laptime gap against the driver who ran on a Soft compound. Because of that, the compound effect is applied in order to level the playing field as if all the drivers performed the lap time on Soft tyres.
The last step is the car effect. Race by race, we are trying to estimate the gap between the cars using our analysis and simulations and we end up with an estimated gap between teams. This gap is removed in order to correct all the lap time and make them comparable to the quickest team.
The resulting lap time is what is used for the metric, having removed all the possible and manageable differences between the teams, track conditions and compounds. Looking at the minimum and the maximum laptime, we can assign a score to the drivers from 0 to 10. Also in this case, the averaging of the results race after race will allow for a more robust result as we get more data.
Conceptually speaking, this is very similar to the qualifying pace. The difference here is that instead of using the quickest lap time, we calculate the average pace of each stint. This is done by removing the slow laps which might be there as a result of strategy or traffic or a difficult warm up phase of the tyre and calculating the average of the core lap times for each stint. Among the available stints, the most representative have been taken into account and among them, the quickest. Additionally, we are also applying the correction mentioned previously such as tyre compound and car performance for the final picture.
This metric is designed to understand who is the driver with the best tyre management, meaning the driver who is able to minimize the degradation throughout the race.
To do this, for every driver and for every stint, the degradation, in terms of seconds per lap lost, is calculated. This is done starting from the lap time, removing the fuel effect as it improves the lap time while the tyres are worsening through degradation, identifying the quickest lap time, and from that lap onwards a linear fit of the lap time is performed. This process allows us to remove the first part of the stint where the lap time without fuel is improving because the tyre degradation is not visible yet or tyres are even improving because of the warm up. We are only interested in evaluating how the driver is able to manage the tyres when the degradation takes effect.
If the best lap time is performed at the end of the stint, then that means the driver didn’t have any tyre degradation to manage and we are not interested in that data point. Among the available stints, for the metric we take into account only the most significant, which means the ones with enough laps, that don’t have many major discrepancies on the lap time, etc. Since different compounds have different intrinsic degradation, a correction is applied, removing the difference between the compounds.
Driver Pit Stop Skill
Looking at the race we always put our attention on how much time is spent on the pit stop, because that can play an important role on the race results. But the pit stop performance is mainly due to the team capability, not the driver. The driver’s job is to simply wait until the team has changed all four tyres.
If we want to evaluate the driver, we need to evaluate what happens in the rest of their time in the pit lane, which is the difference between the total pit lane time and the pit stop: How quick is the driver able to “park” the car for the pit? How quickly are they able to move back to the pitlane post release? This is an aspect never investigated but we decided to track this; while this may not be a game-changing metric (the difference between drivers is small), it’s something to track and see if there are any notable trends or inferences that can be made from this.
We have once again tried to answer the age-old question of how the driver performs once we have removed the all-important effect of the car performance. All F1 fans would like to know who is the fastest driver in qualifying, who is the best on tyre management, or who performs the best in the pit lane situation, for example. With this new insight, we are trying to answer these very questions.