Historical hypotheticals: Part I (Senna, Pryce, Brise)

bellof_monaco

What if Ayrton Senna had survived the crash at Imola 1994? What if Robert Kubica had never entered that fateful rally? The history of Formula 1 is sadly abundant with drivers cut down at or before their peak.

In this five-part series, I will apply the f1metrics model of driver and team performance to simulating historical hypothetical situations. While this exercise could have been performed with the old model I used to generate all-time rankings back in 2013, the new model I have recently developed that incorporates age and experience effects is better suited to the task. It allows us to consider the inevitable effects of age had the careers of Senna or Schumacher been extended, for example. It also allows us to more fairly assess drivers such as Stefan Bellof, who likely had considerable scope to improve at the time they died.

What the model can give us then is a quantitative take on what are usually purely qualitative hypotheticals. Of course, in all these hypotheticals we inevitably face unpredictable factors. So, take these for what they are: a bit of fun.

Method

For each hypothetical below, I simulated either the extension of a driver’s career or a change in teams during their career. Two types of outcomes are presented:

(i) The adjusted driver performance rankings (in ppr) for each season, including the driver’s predicted hypothetical performance at that age and with that level of experience. This is a ranking of absolute driver performances, taking teams out of the equation. In other words, how the model predicts the drivers of that season would have relatively performed in equal machinery. This is the same approach I use for my end of season driver rankings each year.

(ii) The predicted World Drivers’ Championship standings for each season. In cases where these tables are presented, I am including the effects of a driver’s hypothetical team on their points scored. To generate these, I mapped the model’s scoring rate function to all other historical points scoring systems.

Model: I hope to spend some more time describing the model (in mathematical terms) in a future post, but as a brief description to aid interpretation of the hypothetical predictions, this is how it works:

  • The model fits a driver’s performance (points per counting race) in each season as a (nonlinear) function of:
    • Driver performance
    • Team performance
    • Age
    • Experience
    • Random variation in form from season to season
  • Age and experience effects are fit as average effects across all drivers. For previous estimates of approximate age and experience curves, see my previous end-of-year articles for 2015 and 2016, respectively. In other words, the model doesn’t attempt to estimate individual differences in the rate of improvement/decline with age or experience. Although such differences may exist, in general it’s not possible to estimate them, due to many drivers having relatively short careers. Nevertheless, I show the uncertainties in these estimates below.
  • The experience effect considers how many of the past few years the driver was active. Inexperience can therefore apply either to a rookie driver or a driver returning after a long career break from Formula 1.
  • The model is fit to all historical Formula 1 championship events going back to 1950 for all drivers who completed at least three counting races in at least one season, as in previous analyses. Note that as new data come in, the predictions will change (being most sensitive for recent drivers). For example, Sainz was rated the strongest driver in 2017 based on evidence up to that point, but is now ranked only 5th in 2017, given the additional half-season of data from 2018. The predictions presented here are a single snapshot, not a final, definitive statement.

What if Ayrton Senna had survived?

senna-1994

Let’s start with the most popular F1 hypothetical. A google search of the exact phrase “What if (Ayrton) Senna survived” yields over 500 hits.

At the time of his death, Senna was 34 years old, near the peak of his powers. He was facing the best of a new generation: the 25-year-old Michael Schumacher. Which driver would have prevailed in the coming years has been a topic for endless conversation.

This is a nice case for showcasing the model’s age-related decline function. Below, I have plotted Senna’s predicted season-by-season performances, and I have taken it all the way out to 2017 (Senna would then be aged 57). For reference, I have included in each season the best performer, 75th percentile, median, 25th percentile, and worst performer.

senna_hypThe blue curve shows Senna’s actual career. The red curve shows his predicted future performances, with the surrounding shaded error showing the combined standard error in the estimate of Senna’s average performance, the age function, and the experience function.

Senna’s performance is predicted to remain quite stable to age 35 (1995), with modest decline to age 38 (1998), and more rapid decline thereafter. In the years immediately following 1993, Senna is predicted to remain one of the top drivers in the sport, but not able to match Schumacher’s new standard-setting performances. Had he continued to race on, the model predicts Senna would have remained among the top 5-6 drivers until 1999 and the top half of active drivers until 2000 (age 40). By 2005 (age 45), he would be among the weaker drivers on the grid, and in the exceedingly unlikely scenario that he continued racing to 2010 (age 50), he would likely be the clear worst driver on the grid.

In the hypothetical universe where Senna survives Imola, I will assume that he continues racing with Williams to 1997. The first ramification of this is that David Coulthard does not debut with Williams, so the line-up remains Senna-Hill until 1996. With Senna at the team, Jacques Villeneuve is also no longer recruited in 1996. In 1997, when Hill parts ways with Williams, Frentzen lines up alongside Senna.

Of course, with Senna, the butterfly effect is in full motion. Without his death, we may not have seen the sweeping safety changes. Cars may consequently have developed along different lines, and other drivers might have been injured or killed. Senna might also have been in the frame for a move to Ferrari at the time Schumacher did.

Assuming Senna remained with Williams, let’s see the model’s predictions year by year.

senna1994

Projected results for 1994, assuming Ayrton Senna had raced the entire season for Williams in place of David Coulthard and Nigel Mansell.

In 1994, the model sees Schumacher as the clear best performing driver, but Senna narrowly wins the title. This is due to Schumacher missing four of the rounds with disqualifications and suspensions. Would Schumacher have received the same treatment without a comfortable championship lead? Debatable — more butterfly effect.

senna1995In 1995, the model thinks that Senna would have pushed Schumacher much further than Hill managed, but Schumacher’s superb performance is enough to clinch the title: his first in this alternate reality.

senna1996bWith the dominant Williams in 1996, the model sees Senna taking the title quite easily ahead of teammate Hill. Schumacher worked miracles for Ferrari that year, and is the model’s standout performer, but the car was simply nowhere near quick enough to be a championship challenger.

senna1997cIn 1997, Senna again have would enjoyed a significant car advantage over Schumacher in this like-for-like hypothetical — indeed, just enough to clinch the title, in the model’s predictions. Aged 37, Senna would be starting to suffer age-related decline, and his performance is predicted to be only marginally better than Villeneuve’s that year. With a total of six titles (one more than his hero Fangio, and two more than arch-rival Prost) and major regulation changes coming in 1998, including grooved tyres and narrower cars, it’s quite easy to imagine Senna retiring at this point.

Tom Pryce & Tony Brise

tom_pryce_monaco

Tom Pryce was killed in one of the most horrific accidents in F1 history, hitting an unsighted marshal at top speed. Both were killed instantly, Pryce by the marshal’s fire extinguisher striking his head. If you haven’t seen the video footage, I recommend not watching it.

In his three full seasons in the sport with the Shadow team (1974-1976), Pryce was identified as a potential future race winner. Although the car was often unreliable and rarely in the points, it was on occasion quick enough to grab headlines. Across 1974-1976, the team took 3 poles (2 for Jarier, 1 for Pryce) and 3 podiums (1 for Jarier, 2 for Pryce).

Pryce’s results gradually improved relative to Jarier’s across the three years, with the final tally favoring Pryce 20-16 in qualifying, 11-5 in races, and 19-1.5 in points (noting that Jarier’s points tally was hurt by mechanical failures both times he took pole).

With his clear advantage over Jarier in 1976, Pryce was expected to achieve great things in 1977. He was killed in the third race. His replacement at Shadow was Alan Jones, who went on to win the Austrian Grand Prix, then signed for Williams the following year, winning the title for them in 1980.

So what does the model make of Pryce’s potential trajectory?

pryce_prediction2

The best case scenario (top of the shaded area) puts Pryce over the 75th percentile of active drivers, but the more probable prediction is an average performance level. There are three factors counting against Pryce in this hypothetical. First, Jarier (his main teammate) was not a strong driver. Against a driver such as Jones, the model predicts much less favorable results. Second, Pryce was 27 and in his fourth full season in F1 when he died, meaning there was not much further growth potential. Third, Pryce was very much in the win-it-or-bin-it class of drivers. Out of 42 starts, 13 ended in a crash. This makes him one of the most destructive drivers in F1 history.

most_crashes

Drivers with the highest career crash rates (minimum of 10 career starts).

 

Formula One World Championship

Another lost talent from the same era was Tony Brise. He was one of the unfortunate victims of the plane crash that was piloted by and also killed Graham Hill in 1975. Among the first generation to race competitively in karts, Brise was extremely impressive in single-seaters (FF1600, F3, and Formula Atlantic) on his path to Formula 1.

Brise’s Formula 1 career was brief but extraordinarily promising. In his races for Williams and Embassy in 1975, he whitewashed his teammates 9-0 in qualifying, including a 4-0 record against fellow rookie and future WDC Alan Jones. In races, Brise was more of a loose cannon (see his ranking in the crash rate table above), but still beat his teammates 3-2 overall (2-2 against Jones). Jones took a 5th place finish to outscore Brise.

Killed after his rookie season at just 23 years old, Brise had enormous potential to further improve. The model projects a meteoric rise, with Brise most likely becoming one of the sport’s strongest drivers in the late 1970s and early 1980s.

brise_predictionIt’s difficult to meaningfully speculate on Brise’s career path. He was signed to drive for Hill’s Embassy team again in 1976. From there he might have been considered for a Williams drive, like Alan Jones. He might have been in the picture for John Watson’s Brabham drive in 1977. Or he might even have been offered the Ferrari seat when Lauda retired in late 1977, instead of Gilles Villeneuve. In any case, had he landed in a top team, he very likely would have challenged for titles.

 

That brings us to the end of part I of this series. To be continued!

 

Advertisements

25 comments

  1. Ooh five parts, nice. I hoped age and experience effects would bring Schumacher into the realm of mortals, but I guess that won’t happen as long as his teammates are treated as having equal machinery. Other than that Senna’s path seems perfectly believable and not far from what I had in mind. In my ignorance I’ve never heard about Pryce and Brise until recently (or forgot about them when I did). My models seem to think Pryce would be somewhat above average and Brise in a group of drivers just a step below the greats.

  2. Fascinating as always. Schumacher vs Senna would have been great.
    Quick question on the error, is it +/- 2 standard deviations?

  3. Wait, if Senna survived Imola 1994, why is Nigel Mansell listed as having scored 11 points in the alternate championship? Does he drive for another team or does he actually partake in the four races (France, Europe, Japan & Australia) like in the actual real life events?

    Great stuff as always (and Senna doesn’t win 1995 WDC! That’ll wind a few Senna fans up!)

    1. Kevin Lee · · Reply

      To be honest, a more likely (but not in any way mathematical or scientific) scenario would be Mansell not coming back from IndyCar if Senna was still around. It’s not that the former would be intimidated, but the political and economic climate back then would’ve made it unattractive for Mansell to return.

    2. You’re correct, that’s a typo regarding Mansell.

      1. Kevin Lee · ·

        Eh? Do you mean I was right? Mansell’s not going to participate in the ’94 season had Senna was still driving for Williams (which would influence the driver market greatly)?

      2. Yes, you’re correct, i.e., Mansell should not appear in that list and his points should be redistributed. Fortunately, it won’t affect the ordering of other drivers in the list.

  4. I am excited to read the next parts. Very intriguing idea and a great read as always.
    Just one quick question: In your crash rates-statistic, I was missing Andrea de Cesaris. Is he really not in the Top 10? That just seems odd, given his reputation.

    On a side note: I was discussing with a friend, that Tony Brise seems to be a prominent name in the UK, but I (grew up in Austria) was utterly unaware of this particular driver and his history. Is it just ignorance on my part, or is he more of a common name in the UK than in the rest of Europe?

    1. Thanks! Regarding de Cesaris, he had a high crash rate, but doesn’t quite make that list. In 208 starts, I count 38 driver DNFs, or 1 every 5.5 starts. His crash rate is partially reduced by his high mechanical DNF rate (103 mechanical DNFs), although he still wouldn’t top the list if we exclude those.

      My impression is that Brise was quite well known in the UK as a potential great driver, but not yet a superstar.

  5. Great stuff as always! Here’s to Jim Clark, Stirling Moss, Alberto Ascari and Robert Kubica being in future posts.

    1. I have at least three of those planned!

  6. Kevin Lee · · Reply

    Great article, Doc! Shame I missed it during my birthday, when it’s only posted a day before! You should’ve posted more often…

    Also, may I request an updated ‘Greatest Driver of All Time’ post, including this 2018 championship? I think there could be some change in position for N. Rosberg & L. Hamilton, especially compared to F. Alonso’s and S. Vettel’s rankings.

    1. thebest123456 · · Reply

      As far as I read, he has an updated greatest driver of all time post planned, which will also include age and experience effects, but he hasn’t had the time yet.

      1. Thanks for the update. From: Kevin.

  7. Marcelo · · Reply

    Excellent as always!! Does the 1994 forecast consider the real results of the first 3 races? Williams fired 4 WC (Piquet 1988, Mansell 1993, Prost 1994 and Hill 1997). Would have Senna continue in the team in 1995?

    1. Thank you! It does indeed include the actual first 3 race results. Indeed, hard to know how that one would play out.

  8. Great work, as always! Senna’s hyphotesis seems very plausible. I’m a huge fan, but I also believe Schumacher would have been unbeatable in 1995.

    Regarding your next parts, as a brazilian I would like to see Jose Carlos Pace’s hyphotetical career had he survived his plane crash in 1977.

    Thank you again for this piece! Can’t wait for the next part!

  9. Andrew M · · Reply

    Can I ask a question about Schumacher? If he hadn’t come back post retirement, does the new model predict his performances would be better or worse than what he actually performed in 2010 (once you take age related decline into account? My totally subjective impression is that he performed worse than expected, even taking into account his age.

    1. Hi Andrew. The model’s prediction is that he indeed would have performed better had he been active across 2007-2009. The recent lack of F1 experience was a large part of his predicted decline. As it turned out, by the time he had overcome most of that effect by 2012, the age effects were continuing to accumulate, which reduced the improvement in his performance. I’ll run some detailed hypotheticals on this later in the series!

      1. Thank you!

  10. Is there a performance penalty for drivers switching teams? Do they perform worse until they get settled?

    1. Not in the model currently, but something worth investigating! It would be interesting to see if there is such an effect.

  11. Very interesting article as always! But the results are a bit uncertain and doubtful. If Senna survived Imola, he was still 30 points down to Schumacher. Senna-fans are 100% sure he will be champion because Hill came extremely close to Schumacher in 1994. But that’s a myth. The reason there was still a championship battle that year was because Schumacher was suspended for 2 races and disqualified 2 times. Without Senna’s death, the FIA doens’t need an artificial championship, so Schumacher benefited from scoring in 16 races instead of 12.

    The second questionable ‘fact’ is: would Senna stayed with Williams if he doens’t win the 1994 and 1995 championship? The Brazilian doesn’t left McLaren to built up a team, like Schumacher does from 1996 owards. He came to Williams to score instantly. Some people believe Ayrton Senna once wanted to drive for Ferrari. So perhaps he would have left Williams in 1995 after two disappointing seasons?

    If Senna stayed with Williams, I’ll give him the 1996 championship for sure (without considering the regulation changings, following by his death). I’m not sure about 1997. It makes sense this would have been his last season, but Schumacher drove very strong that year.

  12. Leonardo · · Reply

    Any estimates of when we could be seeing part 2? I understand you’re a busy man but am itching to read more!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: