Austrian GP 2018
F1 is heading to the 2nd of the three back-to-back Grand Prix in Red Bull ring. Last year, Valtteri Bottas took the win with Lewis Hamilton only managing a 4th place. Vettel is definitely in need of a win if he wants to keep his chances of challenging for the drivers’ title high.
I think that this will be a race where the top three teams will be pretty evenly matched. The model predicts that Hamilton will take the pole and the race win after his recent strike. I’d bet my money on Vettel or even Bottas for a comeback! Race predictions have been updated after the qualifying session.
Vettel has a 3 place grid penalty while Leclerc has a 5 place penalty.
| Driver | Qualifying Prediction | Actual Qualifying Result | Race Prediction | Actual Race Result |
|---|---|---|---|---|
| Verstappen | 6 | 4 | 5 | 1 |
| Raikkonen | 5 | 3 | 4 | 2 |
| Vettel | 2 | 6 | 3 | 3 |
| Grosjean | 16 | 5 | 7 | 4 |
| Magnussen | 11 | 8 | 8 | 5 |
| Ocon | 8 | 11 | 12 | 6 |
| Perez | 10 | 16 | 14 | 7 |
| Alonso | 12 | 13 | 11 | 8 |
| Leclerc | 13 | 18 | 15 | 9 |
| Ericsson | 18 | 20 | 17 | 10 |
| Gasly | 14 | 12 | 13 | 11 |
| Sainz | 7 | 9 | 9 | 12 |
| Sirotkin | 20 | 17 | 20 | 13 |
| Stroll | 19 | 14 | 19 | 14 |
| Vandoorne | 15 | 15 | 16 | 15 |
| Hamilton | 1 | 2 | 1 | R |
| Hartley | 17 | 19 | 18 | R |
| Ricciardo | 4 | 7 | 6 | R |
| Bottas | 3 | 1 | 2 | R |
| Hulkenberg | 9 | 10 | 10 | R |
Please share the post:
10 thoughts on “Austrian GP 2018”
Is the model able to show the circuit specific tendencies in order to be able to differentiate this list from something which looks much like the current championship standings with some extra weight to the recent results?
Hi Martin,
Yes, the model does take into account the circuit for its predictions. I guess you would expect Bottas a bit higher up, right?
Probably the model did not find this tendency strong enough and made Hamilton and Vettel the favourites.
If you see the betting odds, all bookies say Lewis is the favourite.
I think their models made similar predictions 🙂
If it is about winning money then the trick is to find out where the bookies will be most likely wrong.
In order to do that this model seems to be very (too) conservative and close to an average of recent races. The interesting question is then how input parameters are changing the picture.
Also interesting is to make the predicted standing not dependent on relative driver battles
(since the distribution of positions can be extreme, Verstappen is a good example of a driver who is not likely to finish at place 6 this race, either he does better 60% or he crashes 24.4%, this might make his average position 6, but it is not his most likely place. Only 2 out of the 45 Red Bull races Verstappen finished 6-th. Under normal circumstances Verstappen finishes ahead of Ricciardo and possibly takes a Ferrari or a Mercedes as well. Especially qualifying in 6th place is highly unlikely. Ricciardo is sticking much more around that place and has only his Monaco pole to boost up his score. Last year place six was 3 times Ricciardo 6th, 2 times Raikkonen, and 3 times some lucky driver from a non top-3 constructor)
(An interesting bookie score is on Verstappen to retire. Then you get 4 return, but his chances of retirement are likely to diminish from his 25% if you think about regression to the mean.)
Regarding Hamilton/Vettel/Bottas. Especially if the bookies say that Hamilton is the big favorite then it is a good choice to select the opposite (on top of Hamilton being not so good in Austria, but your model might provide this information more in terms of numbers than my gut feeling). Last race the difference with Vettel was not at all that big. It is just that Vettel did so great in Canada while Hamilton had to get the last bits out of his engine.
Bottas with some hedging with Vettel might be a much more interesting bet for the race winner. Raikkonen has been difficult to guess lately, but given his low popularity among the bookies he is a good risky bet. I imagine that the probability for a podium by Raikkonen, according to your model, are higher than what the bookies believe. Currently he has roughly 3.5 return. Less than 28% probability for him to get on the podium seems a bit harsh. His bad days are not everlasting.
But in any case Bottas should not be predicted to be overtaken during the race.
errors:
The Verstappen retiring comment (and the bookies 4 return) did not make sense. It is opposite and only when you believe that Verstappen probabilities are rising above 25% this is interresting.
“Last year place six was 3 times Ricciardo 6th, 2 times Raikkonen, and 3 times some lucky driver from a non top-3 constructor” I meant ‘this year’
Yeah, all your comments are valid. Still, I’ve not built the model with the goal of winning money so it does not try to produce ‘interesting’ predictions but rather realistic ones “on average”. That’s why the model, as you said, predicts the current championship standings plus some fluctuations based on the driver’s form.
I tried to built models that do not use the pairwise driver battles (actually this model gave the very first predictions of the blog) but switch to this pairwise approach since it gave batter results. I get that Verstappen’s 6th place is not very common but this close to his ‘average’ finishing position.
When I find some time, I intent to build betting-specific models, e.g. a model that directly predicts each driver’s probability of being in the top-10 (I’ll do the same for other markets as well). Then, I’ll be able to compare my probabilities against the bookies odds and select the one with the biggest differences. I can keep you up to date with this if you like to, however I don’t think I’ll have something ready for this season.
Not only for betting but also in general it is interesting for a model to be edgy. If a model turns out to be a vanilla weighted average of the last races then there is not some much to learn from it in comparison to other models.
Your current model already has this capacity, I believe, but it would need to make the effects more salient. How is it that the model makes certain predictions. An example is that when a prediction is made for some specific circuit then you can show how it is in contrast with other circuits.
I have blogged about what are the main factors driving the predictions (see model interpretability) and the circuit is not amongst the top ones. Of course, this is just high level; circuit is probably relevant is some of the predictions but not in all of them.
I agree that I could do a better job at presenting the predictions. I haven’t thought of, or I’m just lazy to, any way to do that while still writing short posts.
Hope the British GP model results are up on the website by July 7 6am GMT
Thanks
If this is about betting.
An interesting bet would be to make a split for Vettel to win and Hamilton on pole.
Currently the returns (bet365) for Hamilton/Vettel are for pole position (1.66/3.25) and for winning the race (1.72/5.00). So somehow the betting odds are more in favor for expecting Hamilton to win than Hamilton to be on pole (a discrepancy that we can take advantage of, since the probabilities for pole position and winning the race are closely linked this season where overtaking is difficult).
How this discrepancy occurs I do not know exactly, but likely the betting agencies expect to get lots of votes for Hamilton to win reducing the return on that bet (to me this makes sense since Hamilton is a favorite based on his past results on the circuit as well that it is his home circuit).
On top of that the probabilities for Hamilton to drop out might be a bit higher since the last double technical failure at Mercedes. This makes it strange that he is given higher winning chances than pole position chances. (although this speculation is risky, while we do not know much about what happens at Mercedes)
Tires for British GP is a factor.
The British Grand Prix is one of three races where Pirelli will bring special tyres with a lower tread depth to help teams keep tyre temperatures from rising too quickly. This has been done in response to the new surface which has been laid at Silverstone, as well as the Circuit de Catalunya in Spain and Paul Ricard in France.
Spain Results
Lewis H
Valtteri Bottas
Max V
Seb V
Daniel R
Keith Mag
Carlos Sainz
Fernando Alonso
Sergio P
Charles LeC
Pole Lewis
French Results
Lewis H
Max V
Kimi R
Daniel R
Seb V
Keith Mag
Valtteri B
Carlos Sainz
Nico H
Charles Hulk
Pole Lewis