Team mate wars, data prediction suggests big shock

During winter testing, many have tried to predict the outcome for Melbourne based on data. The pecking order is still very unclear with Mercedes surprisingly low down on lap times, although the long runs do suggest that things are not that dire for the Silver Arrows.

TJ13 published last week an interesting article using Karun Chandhok’s data-driven predictions, you can read that here. In it, he sees Ferrari ahead, unsurprisingly, but also that Haas should be top of the midfield for the beginning of the season, just like last season.

A fantastic resource for F1 fans interested is a blog called F1Metrics  Mathematical and statistical insights into Formula 1. In it the writer analyses data from lap times, compiles it and predicts or explains the on-track performances.

Earlier this week, F1Metrics revealed some interesting predictions for the coming team-mate battles for the early part of the season based on past form and also data from winter testing in Barcelona.

“In past years, I have used a method to analyze long runs from testing that has proven to be quite accurate for predicting the within-in season pecking order.” claims the writer.

Below is a summary of the process:

  1. I collected stints that were known to be part of a race simulation. Since these stints were strung together by pit-stops only for tyre changes and typically ran close to a full race distance, the fuel loads are known. These stints can be used to anchor the data for stints run at unknown fuel loads. For comparing stints, I estimated the equivalent pace on a full fuel load (70 laps of fuel).
  2. I collected all other long runs for which the tyre compound was known. Stints were cleaned of slow laps (>1 sec/lap slower than laps on either side). I required a minimum of 10 laps with valid times for inclusion in the analysis.
  3. I derived degradation-time curves for each team on each tyre compound. These curves show the relationship between lap time and how hard a driver is pushing (i.e., degrading the tyre compound, as evidenced by worsening lap times). This step is critical, as it allows comparison of stints with variable work rates of tyre compounds. This is the step that is missing from all other online analyses I have found; without it, one may be comparing a stint where a driver was pushing very hard to one where a driver was extracting very little life from the tyres.
  4. Finally, I attempt to estimate the unknown fuel loads by finding estimates that give the best match to the degradation-time curves for each team.

“Of course, even after all this, we may still be comparing cars with differing set-ups, track conditions, and engine modes. But by compiling enough stints, we can hopefully begin to divine the overall trends, and spot outliers. It’s not perfect, but it’s about the best we can do as outsiders, without access the teams’ internal data.” concludes F1Metrics.

F1 team-mate battle predictions

For those who follow F1 and have reached this far in the article probably have realised that for 2019 only two teams have driver pairings that match last year, so really this year is a fantastic opportunity to predict the coming team-mate battle outcomes and F1Metrics tries to do this using their mathematical method.

Follow the link to the model used for the below predictions.

For the analysis, F1Metrics does try and predict rookies by previous form and comparable drivers. The three obvious inclusions to this being Albon, Russell, Giovinazzi and Norris.

“Specifically, I searched for other drivers with similar rookie age (within 5 years), and similar scores on my junior career metrics of achievement (within 30 points) and excitement (within 5 points). I then used their performances to generate a proxy distribution of performances, applying the appropriate corrections for age and experience to match the actual rookie in 2019.” explains F1Metrics.

The predictions

Below is a direct copy of the predictions, the obvious shock probably being Leclerc versus the four time Champion Vettel. But then, think back to 2014 and a rookie driver back then dominated Vettel so who knows!


Let’s start with a controversial prediction! Given his meteoric first season, the model sees Leclerc (with the benefit of additional age and experience) potentially being more than a match for Vettel, with an almost 2 in 3 chance of coming out ahead. I’m personally more circumspect, as I think Ericsson’s uncertain rating may have inflated Leclerc’s status, as I noted last year. In any case, this will be a fascinating match-up.


Hamilton has looked firmly in control at Mercedes since Bottas joined the team. The model sees that scenario as being very likely to continue.


Gasly showed serious promise in 2018. However, facing Verstappen is a tall order for any driver, especially given he is clearly the favored driver within the Red Bull team. The model sees Gasly having about a 1 in 5 chance of outperforming Verstappen in 2019.


The Magnussen vs. Grosjean battle will continue into a third consecutive season at Haas. Grosjean still seems the quicker driver when on form, but his erratic performances across a season tend to always cost significant points. The model sees Magnussen as the slight favorite here.


This is the most one-sided prediction on the grid. There are two factors contributing to this: (i) although the model’s ranking of Sainz was downgraded last season, it still remains relatively high; (ii) historically, rookies are very unlikely to outscore their teammates. Norris, who was a very talented junior, surely rates his chances higher than this.


This one is surely the most mouthwatering new pairing on the grid. How does Hulkenberg measure up to Ricciardo, a driver who has a strong claim to being among the  sport’s elite? As the model sees it, this one could easily go either way!


Experience vs. youth should make for an intriguing battle at Alfa Romeo. The model views Raikkonen as the favorite, with a 1 in 4 chance of Giovinazzi prevailing over Raikkonen.


Albon and Kvyat are both receiving recalls from the Red Bull program to drive at Toro Rosso. Kvyat has more F1 experience and undoubtedly the better junior career, but has he mentally recovered from his previous demotion and sacking? If he has, expect Kvyat to take the lead at Toro Rosso.


The experienced and very capable Perez is likely to give Stroll a much sterner test than Sirotkin did last year. The model sees Stroll as having a 1 in 3 chance of outperforming Perez; that number is probably inflated due to Stroll’s large rating gain from the result of Baku 2017, as I noted last year. Perez is surely the favorite in this one and the match-up should help to clarify Stroll’s actual level.


Kubica, on comeback, will face Russell, the reigning F2 champion. Under the model’s assumption of no effect of Kubica’s injury, it sees Kubica’s age and talent counting for enough to come out ahead. In reality, we will have to see if the injury matters.






Join other TJ13 readers in the new F1 fantasy league: Code – 9f818a5204




3 responses to “Team mate wars, data prediction suggests big shock

  1. As there are no changes to the engine regulations, I think it is pretty safe to say that Mercedes will be right up there again with their far more powerful and superior PU. Which all means that Hamilton only has to beat his ‘wingman’ (90% chance) to bring home another championship.

  2. Pingback: How to predict the winner of the Bahrain GP - thejudge13·

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.