and so it continues……

Posted on 16 April, 201816 April, 2018 by Nicholas Kell

It’s always good to have 3 races worth of data in the bag. I think this is the first year in a long time where on the surface the season feels very different, well, if the media are to be believed.

It’s is different, but not so much as one would think when looking deeper into the data. Now the Value charts are available, including the breakdown by team, things are probably a little clearer. I think in summary the only big changes are a swap in fortunes between McLaren and Force India. The other shocker (although not that shocking when last seasons symptoms are taken into account) is the current Value of Williams. All this emboldens me to extend the model further to look deeper into the team dynamics and ‘casting’. I won’t burden this blog with my thoughts, but suffice to say when you have a car powered by Mercedes and designed by Paddy Lowe, where else can you look for the current results? Could the same be said for Force India? At least in that team we can make some year on year comparisons with the people? But further can it explain McLaren’s results too? They can’t blame the engine anymore and you can’t help feeling that they should be doing better than they are?

I have published the current team Value analysis charts on the basis of last years end positions (which is normal practice). However, ahead of the next race I will have to account for the change in fortunes for McLaren and Force India, it will just not be helpful to continue to publish them on the current assessed predictions.

So that just leaves with what is happening at the top? It is certainly closer than it has been, but I am not sure by that much. Whilst the data suggests Ferrari are in the lead by a very small margin, their trend is downwards on a fairly steady trajectory. Mercedes on the other hand are slowly rising race on race, which suggests on balance Mercedes will win overall, but it will be close.

So the readjusted places and targets for the remains of this half of the season (which really describes what the season should look like, a close flight going on for 1/2, closely followed by a one horse race for third, 2 horse races for 4th and 5th with a large gap back to the rest):

1st with 2nd for Reference – Ferrari

2nd aiming for 1st – Mercedes

3rd aiming for 2nd – Red Bull

4th aiming for 3rd – McLaren and Renault

5th aiming for 4th – Torro Rosso and Haas

Rest aiming for 5th – Sauber, Williams and Force India

And finally my end of season predictions having mulled the data over:

1st – Mercedes (just, it will be close)

2nd – Ferrari (see above)

3rd – Red Bull (confident, and closer to Ferrari than last year)

4th – Renault (confident)

5th – McLaren (confident)

6th – Haas (confident)

7th – Torro Rosso (confident)

8th – Force India (confident)

9th – Sauber (close to)

10th – Williams (just)

A great new season for data

Posted on 9 April, 201811 April, 2018 by Nicholas Kell

As always it is too soon to make any sort of predictions/analysis worth quoting but looking at the numbers after Race 2 the 2018 season is already looking much different to previous years, more so in the middle runners.

A quick look at the ratio spread show that there is more evenness, but for how long this will last time will tell. However it is a good sign of things to come.

I m not going to focus on the matters which have swamped the main media, ie. Ferrari vs. Mercedes, for me the telling signs in the data are McLaren’s current position and Williams’ demise?

I think China will start revealing the season’s secrets and here we will start publishing the all important Value charts.

2017 Change?

Posted on 27 April, 201727 April, 2017 by Nicholas Kell

As is the tradition, I have kept my thoughts internal until at least 3-4 races in, but I now feel ready to say something meaningful ahead of the 4th race.

As you may have seen some new charts have been constructed over the winter break. The automation in R and the API to Plotly is very efficient. We seem to be only held back by analytical ideas now. The top page of the White Paper now has a team performance ratio breakdown which tells the story of the whole season in one go quite nicely.

On the Team Value pages we also have a nice visualisation of how closely matched the drivers are in the teams. One to watch as the pundits speak through the season as sometimes things aren’t always what they seem.

So, I guess finally it may be worth a few words on what we have witnessed so far and whether the new 2017 regulations have really changed anything?

Whilst the Driver’s competition has a different dynamic this season (how long that will remain who knows), I am yet to be convinced how much will change for the teams as the season progresses. Ferrari have no doubt improved the engineering offering but we have yet to see Red Bull at full tilt yet. The most interesting part of the Team competition is probably in the 4-5th place area. So far Force India are being very consistent and it would be great to see them maintain that position again.

2016 – End of Season Thoughts

Posted on 7 December, 20167 December, 2016 by Nicholas Kell

I was going to write my season wrap up last week but I am glad I held off as I would have ended up having to write a supplementary post anyway now that Rosberg has announced his surprise retirement.

Reflections on Model

Overall I am really pleased with the 2016 model work. It is as simple as it can be but now scalable and pretty robust going forwards. More importantly I reckon it was capable in its current design of predicting the top 3 teams for the Constructor’s Title from about mid season. The one design adjustment that had tom be implemented was coping with drivers switching teams mid-season. But the work done to address this has made the preparations for 2017 Model a little easier as the thinking has now been done. Whilst this model was not required to predict that Mercedes were going to win from the off, it was able to dismiss the Ferrari hype far earlier than most commentary out there. Further, whilst we did not need a model to predict that Red Bull would have a much better season, the model indicated far earlier than wider commentary did that they would beat Ferrari but of particular please to me was that it told us that Force India would beat Williams.

Implications for 2017

OK, so I guess a lot are asking how the regulation changes for 2017 will impact the model? Well, I am not sure it will because the model is still focussed on some pretty atomic data and fairly basic logic which is mutually exclusive to the regulations itself, that is the beauty of it. This enables the model to be run meaningfully on F1 data back to the 1950s if required. However, it is probably only meaningful back to the 90s when most of the key players of today were just starting out. And that is the clue to how we will move the model on into next year. The drive swop in 2016 forced us to change the model to account for drivers not being tied to a team for a whole season (and thus making the model simpler). We had to add a ‘team’ factor to each set of race results for each driver. Now that that is done it is fairly easy to scale this to take into account other key personalities in the team’s make up and results. We will start simply with just the Chief Engineers and Strategists and see where it goes. The key thing is that we will be able to calculate a new team coefficient and see how key individuals score when working with others. F1 is a small world and the results and thus these individual’s impact on the sport as they have rode the merry go round are all available back to the 90s so we should be able to see some trends.

Again you don’t need a model to know that Ross Brawn and his magic worked pretty much wherever he went regardless of budget. Ferrari are currently the antithesis of this in 2016. However, F1 will continue to move personnel around the teams and we will hopefully be able to assess in advance the impact of these HR decisions in advance through next year’s model.

Of course, Rosberg’s untimely departure has beaten me to the work! However, I am not convinced that driver selection is the main route to success that many believe. Again Ross Brawn’s profile is indicative of this assumption. It is of course great media fodder for the off-season and it will certainly impact Mercedes somewhat going into 2017. However, I think the key position’s at Mercedes already are a winning formula and Mercedes governance will restrict much of the speculation. If I had to predict? Probably Bottas or Werhlein. I cannot see at this stage how deals can be done outside of this? And knowing the German decision making I think they would probably like to make the Bottas option work.

Are the Numbers Working?

Posted on 6 October, 20166 October, 2016 by Nicholas Kell

We are well through the season and the final results are pretty tight to call. My standing questions still remain that the model tries to address and they are:

will the team final standings settle on the mid-season predictions?
how early in the season does the model give signals to the overall competitiveness of a team?

Before I give some thought to the two bullets, let’s just remind ourselves of the conceptual approach to the model that has been used this season to calculate the ‘Value‘ Coefficient. At the beginning of the season we nominally placed the teams in a predicted end of season order and then calculated the race on race Value target that each race Value Coefficient could be measured against. We could have wasted much time trying to work this out but I simplified the matter by just placing them in the order they finished the last season knowing that mid-season we could reflect on matters, re-order if necessary and then recalculate the respective targets and let it run to the season end.

Fairly early on in the season some very obvious signals were emerging which suggested the following:

Red Bull and Force India were the only teams meeting or exceeding their Value Targets. They were originally predicted to be placed 3rd and 5th respectively come season end.
Ferrari and Williams were consistently well under performing their Value Targets. They were originally predicted to be placed 2nd and 3rd respectively.

I was sufficiently emboldened by the time Verstappen moved to Red Bull for me to place a £50 wager on Red Bull coming 2nd in the Championship, which now seems to make more sense now than it may have back in the season!

Then came the summer break and the review. I did the following:

moved Red Bull to 2nd and Ferrari to 3rd
moved Force India to 4th and Williams to 5th
recalculated the Value Target Values
monitor as the season progressed

If I had to call it now I would say the following with some confidence:

Mercedes, Red Bull, Ferrari, Force India, Williams, McLaren…….

Why?

Mercedes does not need further explanation the only question that remains is which driver will win and that is not really the main objective of this model.

Red Bull have not only had to be moved up in the model, but looking at the results on the revised targets indicates they are more than consistently achieving a 2nd place performance. I would go further, I do not think it will even be close to Ferrari!

Ferrari despite all the hype and heritage it is clear now (as is probably has been most of the season) that they are not best of the rest. I think the ‘Casting’ phenomena will be well explored in this team going into next season. I am convinced that the car is good but that it is the mix of key individuals that has explained the overall result for this team.

Force India are not quite on top of their well deserved current 4th place yet, but I would be very surprised if this was lost by season end. This team showed glimmers of this level of performance last year but didn’t quite beat Williams as a result, but this year has clearly been different. It is so good to see a team like Force India perform so well on the budget they have and their 2nd place in our Value Chart is no surprise.

Williams always promise so much of late and have just not delivered. Like Ferrari the ‘Casting’ factor will reveal much going into next year. It is hard to believe that this team was best of the rest only a few seasons ago. What has gone wrong? Is it the car, the drivers, the senior leadership or a combination of all these things. They must be asking the same questions. Should Force India finish ahead of them come season end then it will be interesting to compare ‘Casting’ analysis as it will not be money that delivered the difference.

Look Ahead to 2017

My next blog will look at what we are prototyping for next year to move the model onwards. I will introduce the ‘Casting’ factor to the Value Coefficient which I hope will be the final step towards really understanding how the teams performance is influenced by the mix of individuals in key team posts over time.

2016 Mid-Season Thoughts

Posted on 15 August, 201615 August, 2016 by Nicholas Kell

With Spa only a week or so away and the data for the start of the season now well processed and understood it is worthy or a few words and maybe even a venture into predictions!

For me the real story throughout 2016 was that of Red Bull. I had felt early on they were looking like the number 2 team and as you can see from the Value analysis below they are pretty much there.

Red Bull Analysis 2016 Mid Season

It is mid season and I have reprojected all the target plots on the basis of revised end predictions as Red Bull were the only team that were exceeding their start season predicted finish. Therefore I recalculated on the basis of Red Bull finishing 2nd. I have also introduced a new plot too which I hope does not complicate the Value interpretation but feedback is always welcome. The extra plot is the target for the next place in the table so in the case of the Red Bull plot the dotted line is where they need to aim for 1st place. I have done it not because they could seriously win, but it provides some real context for their progress against the 2nd place line.

The image also shows the effect temporally of the switch of drivers following the Russian race. Despite Verstappen’s performance in Monaco, he has added nothing but value overall. I guess the guys at Red Bull have crystal balls?

Of course the move of Red Bull to 2nd is at the expense of Ferrari to 3rd. I have not published the Ferrari Value chart with the extra plot, but just with the reprojection you can see the context for their fall from grace:

Analysis of Team Ferrari Value Coefficient

They are only just hanging on to the 3rd place!

Beyond the top 3 there is a story worth following in the 4th place battle; that of Force India and Williams. Williams have had a miserable season and I reckon Force India will be taking that 4th place either at Spa or not long after. What a great result for Force India but it does beg the question as to what Williams need to do to improve?

Some 2014 Retro-Analysis going into 2016…..

Posted on 4 March, 20167 March, 2016 by Nicholas Kell

It would be easy in all the excitement this week to have overlooked a curious story in The Mirror.

Blink and you may have missed it in all the squabbling over Qualifying Formats.

It immediately reminded me of the purpose of the data analysis model here at this website, ie. to use the freely available data to better understand F1 Strategic Decision Making. It is easy to get distracted with the easy tasks of taking the latest data, adding it to the machine learning model an produce some charts which make the model easier to understand whilst at the same time forget that the whole reason we are going to this bother is so we can use it to better understand/preempt F1 Team decisions.

Back in 2014 when this work was at its embryonic state, we wanted to concentrate first on establishing a solid Value model for the teams as a whole. Without a Value model, it was going to be difficult to understand how the drivers contributed to the Value as a result of their efforts and thus how the Teams would retain/trade them. Because of the skew of effort, we merely took a punt at the driver trading that was taking place based on experience and the primitive figures the Value model was producing at the time. The curious thing when I saw this article was it backed up some of the hypotheses we were notionally running at the time. The data and the observed Performance and thus Value of Alonso suggested he was on the move, the questions was where to? We were not convinced that Mercedes would do a swap deal, although Hamilton despite being on winning form was not happy with the way things had been managed at times. When you are an asset like Hamilton who would not be tempted by Ferrari if under that pressure? It is of course interesting that this story has emerged now and from Alonso!

Focussing back on Team Strategy it was all credit to Totto Wolf and probably Niki Lauder more so that the drivers were mentored back to the Mercedes’ main effort, something other teams should take note of.

It has also given me pointers for this years work behind the scenes on this White Paper. Building on the solid foundation of the Value model, we will now start looking at historic and current drivers contracts and associated commercial outcomes balanced against the Value model. More to follow in later blogs.