A Teammate-Based Open Wheel Driver Rating

by Sean Wrona

Last December during the auto racing offseason, I had the idea to conduct my first serious attempt at an objective driver rating. What primarily drove this was that I was curious whether Formula E, IndyCar, or Super Formula had the deepest field. While everybody knows Formula 1 is the most prestigious open wheel series in the world, there doesn't seem to be much of a consensus on which of those series is in second place. Some would argue Formula E because it is a global series where most of the best drivers to lose their Formula 1 rides eventually end up. Some would argue IndyCar because of its mix of ovals and road courses and street courses. Some would argue Super Formula because their cars come closest to matching F1 cars in performance. Needless to say, I was not remotely sure which of these series was actually the most prestigious and I wanted to have a better intrinsic method of this in case I want to do another global driver ranking some year, and I also thought I needed to do this to give me some context that would allow me to create a more accurate top 1000 drivers list if/when I do that after the completion of my typing book (the first draft of my manuscript for that is two chapters away from being finished.) Needless to say, I was also curious about how the best drivers in each of the other series would do in F1 and also how each driver across all these series would do adjusting for equipment, so I created a very simple model that should help answer all these questions.

This model builds on my earlier research wherein I calculated all the Cup and IndyCar drivers' records versus their teammates. Just as in those articles, I determined teammate records by counting the number of times each teammate beat the other in the races where both drivers finished. Although other analysts like F1metrics think that one teammate wins by default if the other teammate crashes, I decided not to do that because crash avoidance does involve skill but it also has a luck element that can at times be difficult to separate. Additionally, if I want a realistic model that best reflects how all drivers would do in the same equipment, I want to remove any team-related factors as best I can. It stands to reason that crashes are often caused by mechanical failures and teams that are poorer and/or have slower cars are probably more likely to have mechanical failures and therefore crashes that are not the driver's fault. To make this most accurate, I think only the races where both drivers finish should be considered. I also did this in part for convenience sake because unlike F1metrics who analyzed Formula 1 races only for the most part, I included results from many, many series in my analysis and while it is usually easy to find the races in which a driver retired, it can be very hard especially as you go into the distant past to find the DNF/retirement statuses, particularly for minor league racing series.

For this analysis, I initially intended to include all the drivers who made at least one start in any of the four major league open wheel series (Formula 1, Formula E, IndyCar, and Super Formula) from 2010-2020, but then I got more ambitious and I expanded my scope several times. I decided ultimately to include any driver who made a start in any of those series since 2000. However, I did not merely include results from those series. I strove to calculate teammate records for basically every open wheel racing series in the world that I could as long as it involved drivers eligible for the list. I decided that what matters in drivers' minor league careers is how they perform against other drivers who also become major league drivers. My goal here was to create a ranking wherein the most average open wheel drivers at the major league level would have a rating of 0. To do this, I decided to entirely ignore any drivers who did not make the "majors" as obviously they will on average be wore than the drivers who did and I didn't want to inflate drivers who dominated a bunch of nobodies in minor league racing series.

I briefly considered including a few other series such as A1GP and Superleague Formula as major leagues because I know they were highly regarded in their time, but I decided not to because they were extremely short-lived, did not really become "destination series" for their drivers like the other four did, and because driver lineups tended to change from race-to-race to an absurd degree in those series, but I did include results in those series in the analysis. So if two drivers who eventually made major league starts competed against each other in anything: GP2, GP3, British Formula 3, German Formula 3, Japanese Formula 3, the Macau Grand Prix, the Masters of F3, Indy Lights, etc... I included it if I could find the data. All races regardless of the level were given equal weight in the analysis, so there are some drivers who will be overrated if they had a much stronger junior career than major league career, which is one of the main limitations of the model, but I still like the results. There were a few series I excluded such as the Formula Two championship from 2009-2012 and other series where all drivers theoretically had the same equipment. I probably should have included them but since there were no teams or teammates per se I wasn't sure whether that would skew the results in some way; not that many major drivers competed in those series anyway so it probably wouldn't affect much. I'm probably going to be a hypocrite when I repeat this analysis for stock car racing next because I do think I'm going to include IROC results.

For the most part in this analysis, I only included races that were a part of an official points championship, but there were a few exceptions. The Macau Grand Prix, Masters of F3, and Pau Grand Prix races are historically among the most important minor league races where drivers from all over the world come together and compete and there are so many major crossovers between teammates in those races that I would be remiss to exclude them. I also decided to count the Nikon Indy 300 IndyCar race from 2008 and the non-points Fuji events from Super Formula. The IndyCar Surfer's Paradise was only not added to the schedule for political reasons and it was taken just as seriously as any of the points events so I thought I should include it (but I did not include any of the races that were intended to be completed but abandoned mid-race like the Charlotte race in 1999 or the Las Vegas race in 2011, nor did I include results from CART's short-lived all-star event.) I also included the non-points Fuji races in Super Formula because similarly to the IndyCar Surfers Paradise race, the competitors took them just as seriously as the points races and I felt they provided too much information to discount them, especially because Super Formula doesn't usually have very many races in a year.

Originally I planned to only include races from 2000 to present, but I eventually realized that creates an extremely skewed perspective of certain driver's careers, particularly those who were still active in the 2000s but whose real heyday was in the '80s or '90s. Ultimately I decided to cover the entire careers of all drivers who made a start since 2000, even including their '90s, '80s, or '70s races, and even including teammates who did not make a start since 2000. So Eddie Cheever's rating considers how he did against Alain Prost and Michael Andretti's rating considers how he did against Ayrton Senna, and so on. Even though Prost and Senna were not eligible for the list themselves, I did count them and many other '90s drivers because they had teammates who were. I eventually decided to add a bunch of these drivers to the analysis as well. If a driver did not make an open wheel start after 1999, I decided to include them if all their teammates did, or even if all their teammates' teammates did (either way, I already had all of these drivers' teammate records calculated, so why not?) This enabled me to include a lot of '90s drivers who did not compete after 1999, such as Greg Moore, Gonzalo Rodriguez, Gerhard Berger, Scott Pruett, Willy T. Ribbs, Hiro Matsushita, and so on. Even Emilio de Villota ended up being eligible because his only teammate relationship was in the short-lived British Formula 1 series against Eliseo Salazar... in 1980! This pretty much enabled me to cover every important driver since 1996, a pivotal turning point in open wheel racing history since that was the year IndyCar split, the year Michael Schumacher started at Ferrari, and the year Japanese Formula 3000 was renamed Formula Nippon and really came into its own as a major league racing series in Japan. However, I still counted all Japanese F2, Japanese F3000, and Formula Nippon results as Super Formula races regardless of what the series was called. Similarly, I counted all IndyCar points races regardless of what the series was called at that point, so I include late-period Champ Car, early IRL, and so on towards the IndyCar statistics. And like I said, I also threw in that one Surfers Paradise race, but I think that is pretty justified.

I obtained results from myriad different sources. Wikipedia was probably my main source because for almost every current racing series in the world, they have their nice color-coated graphs containing all a season's results for each driver with the DNFs conveniently colored in purple, although if there were conflicts between the data listed on a Wikipedia team or driver page and a season page, I did check to make sure the data were right whenever I caught this (and sometimes, especially in CART, the DNFs weren't proper recorded when a driver scored points.) I also used racing-reference, Ultimate Racing History, the F2 Register very often, and I used a handful of other sources here and there when other sources failed me. Sometimes I had to go on the French Wikipedia to find earlier Macau Grand Prix results as weirdly there were pages for many of them without any results sections in English. I looked through each driver's "What links here" page on Wikipedia to make sure I wasn't missing any series or races that I wanted to include. I was very thorough.

Although the model is indeed very simple, it remains probably the most rigorous thing I have ever done on this site because I decided to conduct this analysis in the way that real sports statisticians do, as this model is similar to a lot of rankings systems in "stick and ball sports." I started out by taking the difference between a driver's record against each particular teammate and that teammate's record. For example if Michael Schumacher had a 12-3 record against Rubens Barrichello in the races they finished and Barrichello had a .400 career record against all his teammates, Schumacher would score .200 in that particular year against Barrichello. Why? Because if Barrichello had a .400 career teammate record, that means that the average Barrichello teammate would have a .600 record against him while Schumacher had an .800 record, and .800 - .600 = .200. I then calculate this across each driver's entire career for all their teammates who made a start in any of the four major league series, even if it was in a minor league and even if that driver did not make any major league starts after 1999.

To show a simple example of how it works, I provide Greg Moore as an example. Moore ended up ranking first on my initial driver ratings list among all drivers with 5 or more starts, but because his career sadly was so short, he had a small enough sample size to make it easier to see how this calculation works over an entire career:

Moore had only one major league teammate in his Indy Lights career and that was Claude Bourbonnais, who only made a small handful of races in CART and the IRL and didn't finish any of them, but was probably one of the most underrated drivers of the '90s regardless. Two years prior, Bourbonnais had significantly outperformed future CART and F1 champion Jacques Villeneuve in 1993 in the Toyota Atlantic series and beat him 5-2. Mind you, this was a mere 2 years before Villeneuve's CART title and Indy 500 win and 4 years before his F1 championship. 2 years later, Bourbonnais progressed to Indy Lights and Moore swept him 3-0. That gave Bourbonnais a career record of .500 and gave Moore a rating of .500 against Bourbonnais because he had a perfect record against a driver he was expected to only perform .500 against. That came in a year when Moore had the best Indy Lights season in history, winning 10 of 12 races (while Bourbonnais wasn't even full time.) Moore did not have a teammate in 1996 and 1997, but his Forsythe Racing team expanded to two cars in 1998 and he again scored a perfect record against his teammate Patrick Carpentier, who himself had a surprisingly good record of 58-41 (.586.) This means that the average driver would be expected to beat Carpentier 41.4% of the time, but Moore beat him 100% of the time giving him a rating of .586 against Carpentier that season. The next year, Forsythe added a third car for Tony Kanaan and Moore went 4-2 against Carpentier and 6-2 against Kanaan. Since Moore's record against Carpentier was 1/3 less than the previous year, so his rating dropped from .586 to .253, which is still extremely impressive mind you. But against Kanaan, he was even better and Kanaan had an even higher career teammate winning percentage of 354-185 (34.3%.) Since Moore beat Kanaan 75% of the time and Kanaan's teammates on average bat him 34.3% of the time, that gives Moore an astonishing rating against Kanaan at .407. What makes this even more impressive is that Moore had the inferior Mercedes engine that year while Kanaan got to keep his superior Honda from the previous year yet Moore still outperformed him that badly. Kanaan and Carpentier even at that point were no slouches. Moore was the only full-time teammate in Carpentier's entire career who had a winning record against him and Kanaan only lost to three teammates in his entire career: Moore and Carpentier were two of them before he lost to Scott Dixon in his decline years. Based on this, a strong case can be made Moore was actually the best driver in CART in those years but his equipment let him down (it certainly wasn't Alex Zanardi, the fellow 1996 CART rookie who drastically overshadowed him but had nowhere near that impressive of a teammate record. Let's put it this way: Moore blew out Kanaan by about as much as Kanaan blew out Zanardi two years later.) To calculate Moore's career average, I simply average all these ratings, weighted by the number of races. In this case it would be: (.500*3 + .586*6 + .263*6 + .407*8)/23 = .42565 (although it appeared to reduce to .425 in my database due to rounding error.) This initial rating was so impressive it was actually blowing out Formula One champions, and even after I reiterated, only one F1 champion actually passed him.

However, obviously this alone does not create any kind of realistic ranking. It's a good start, but certain drivers will be significantly underrated because they had much stronger teammates than average and others will be overrated because they had weaker ones. Additionally, many clearly inferior drivers have better teammate records than clearly superior ones. For example, Dario Franchitti has a career teammate record of 169-157 while Sage Karam's is 39-30 and Santino Ferrucci's is 49-30. Technically both of them have better teammate records because they had substantially weaker teammates, so the initial calculations mean that beating Ferrucci is worth substantially more than beating Franchitti, which doesn't make any sense. Hence I decided I needed to reiterate the data by plugging in the driver ratings from the first iteration and replacing each driver's record with their initial rating. Since Franchitti had a higher rating than Ferrucci and Karam, beating Franchitti becomes more and more valuable in future iterations. Notice that some drivers with records deep in the negatives actually had positive ratings because they had exceptionally strong teammates, such as Rubens Barrichello and Hélio Castroneves. That will in turn lift up all their teammates, which will lead to a cycle that will also lift up them in future iterations. Meanwhile, a lot of drivers whose best results were in minor leagues against relatively minor drivers or competed in shallow periods in major league series like the early IRL or late period CCWS will go down. This iteration process is essentially the same as how you will see people calculate the "strength of schedule" in other sports and then continue to reiterate so that you get the strength of schedule when considering teammates' strength of schedules when considering teammates' strength of schedules in an endless recursion process until the data converges. I repeated the iteration 30 times, at which point I felt the data were close to converging. I will again use Moore as an example here:

Here instead of the teammate records, I plug in the actual teammate ratings. I previously calculated Moore's rating as .426 above (although it is listed as .425 in the table below.) Bourbonnais's initial rating was .073 (an above average rating makes sense since he had a .500 record against Greg Moore and Jacques Villeneuve and no other teammates) while Carpentier and Kanaan were curiously tied at .105 on the first iteration, and they would remain very close on all subsequent iterations. Because the model was designed to have an average rating of 0, that means generally the scores converged between the range of .5 and -.5 for all drivers, so I chose to use this range of 1 unit centered around 0 to plug in for the teammate records. A rating of >= .5 was counted as an undefeated record, while a rating of <= .5 was counted as a winless record. I chose to do that because I didn't want to reward the drivers who lost to drivers with a rating >= .5 or penalize drivers with a rating <= .5. If a driver loses to a driver rated >= .5 on any iteration or beats a driver rated <= .5, it simply counts as a zero and does not affect that driver otherwise. Because Moore, Carpentier, and Kanaan all had exceptionally strong records against their teammates with none of them losing to any other teammates outside that triumvirate except for Kanaan losing to Scott Dixon late in his career, they all dragged each other higher and higher on later iterations, and the same applied for all drivers who had exceptionally strong teammates. So it makes sense that the Formula One drivers generally gained the most across numerous iterations, but that doesn't mean lots of IndyCar and Formula E drivers didn't also gain in the reiteration process (the Super Formula drivers, not as much.)

It was kind of shocking to me that Patrick Carpentier never lost to a teammate except Moore, not even Paul Tracy (who he tied 10-10, and Carpentier was Tracy's teammate during his championship season), but that does clue me in to one of the biggest weaknesses of this model. Because it merely reflects which teammate beat the other, it does not consider the actual result or how far apart the two teammates were in a race. Carpentier may have tied Tracy in the teammate record, but in terms of wins, Tracy beat Carpentier 9-2 and obviously utterly dominated him (Carpentier only finished 5th in points in Tracy's championship season.) Winning while your teammate finishes tenth is more of a significant defeat than finishing fifth while your teammate finishes sixth, and this does not reflect that at all, which does create some weird skews like this. Carpentier tying Tracy in the teammate record is probably one reason why he and his important linked teammates like Moore and Alex Tagliani come out substantially higher than they should, but there are still relatively few cases of this overall. Usually if one teammate dominates in terms of teammate head-to-heads they also dominate in all measures of performance, and situations like this or Graham Rahal and Takuma Sato (where Rahal finishes higher than Sato the vast majority of the time but contends for race wins far less often) are rare. Regardless, this does indicate to me that I was significantly underrating Carpentier, who I basically viewed as the Gen X version of James Hinchcliffe but was apparently substantially better - even if you adjust for the fact that he didn't win as much as his teammate performance would indicate, he still probably should be rated over Hinchcliffe in any serious ranking.

Below I list all the eligible drivers who had 5 or more single-race teammate comparisons. This includes all drivers who made a start in Formula 1, Formula E, IndyCar, and/or Super Formula since the year 2000 (with the exception of a few drivers like Dillon Battistini, Stanton Barrett, and Koki Saga who had fewer than five "teammate-races" against major league teammates, usually none.) I also include all drivers whose teammates or teammates' teammates all made a start since 2000 even if they did not, which leads to a lot of exclusively '90s drivers appearing here as earlier mentioned. Unlike some of my less "scientific" rankings, this time you will notice some bizarre outliers, which should prove that this is indeed a mathematical model and not some list I decided upon myself.

Foremost among these is Mark Taylor, the 2003 Indy Lights champion, whose career effectively ended after a very disappointing 2004 rookie campaign. He ranks second among all drivers with a significant number of starts despite all his relevant races being in one minor league season (the 2002 British Formula 3 season.) In that year, he beat two other drivers with similarly very short-lived major league drivers: IndyCar driver Richard Antinucci and Champ Car driver Ronnie Bremer. Taylor beat Antinucci 14-1 and Bremer 11-3 in that year of British F3 which was enough to give him an astonishing rating of .387 because Antinucci and Bremer both also had inexplicably high ratings. To their credit, Antinucci did have a winning record in the minor leagues against two F1 drivers (Kazuki Nakajima and Sakon Yamamoto) and Bremer had an undefeated 5-0 record in Champ Car against Bjorn Wirdheim and Ricardo Sperafico, but none of them really did anything against any drivers who had long-term relevance (aside from Antinucci's record against Nakajima, who did go on to win a Super Formula championship.) These three drivers are extremely dependent on each other as 45 of Bremer's 52 "teammate-races" were against Taylor and Antinucci, all of Taylor's were against the other two, and 46 of Antinucci's 102 races were against Bremer and Taylor. These three drivers were so dependent on each other that they sort of accidentally raised each other up to what I believe are absurdly overrated driver ratings, especially on later iterations when Taylor becomes the highest-rated driver in the entire model. This suggests maybe I should throw out all drivers who did not have two full seasons in any major league racing series (which none of those three drivers did.) Unlike in the cases of Claude Bourbonnais and Patrick Carpentier where I believe I discovered a genuine underratedness, I would not say the same thing for these three drivers.

Some of the other surprises really aren't if you look more carefully at their teammate records. Gonzalo Rodríguez, the other CART driver to die in 1999, is right behind Moore on the initial model because he had undefeated 5-0 record that included a 2-0 sweep of Justin Wilson in Formula 3000 and a 1-0 record of Al Unser, Jr. in his brief CART career. This rating is clearly not spurious but 5 races is obviously a small sample size. Franck Montagny had a winning record against every driver he ever competed against except Takuma Sato in his brief F1 career: this includes an astonishing 17-1 record against future 4-time Champ Car champion Sébastien Bourdais, an 11-1 record against Heikki Kovalainen (beating him substantially worse than Lewis Hamilton did), a 9-6 record against Justin Wilson and a 2-0 record on his IndyCar debut in the Champ Car finale where he beat Paul Tracy; his rating of .358 is clearly not spurious. He is genuinely a driver who deserved a full-time major league open wheel ride at some point and never got it until the Formula E season, where he likewise finished second on his debut before his career effectively ended when he failed a drug test and was replaced by Jean-Eric Vergne, who went on to win multiple championships. Montagny was clearly a great driver not recognized as such (except in sports cars.) James Courtney dominated 2010s legend André Lotterer in British Formula 3 before eventually winning a V8 Supercars championship. Darren Manning was one of only two teammates to beat Scott Dixon in his career and also beat Rodolfo Lavin as a rookie worse than Patrick Carpentier and Paul Tracy did the next year; Manning also gave A.J. Foyt's team two of their best three points finishes since the CART drivers and teams entered the IRL, but those seasons didn't count because he didn't have a teammate. Rinus VeeKay looks absurdly high on the surface, but his 29-2 record to date is astonishing, particularly his 2020 rookie season when he had an undefeated record against both of his much more experienced teammates (7-0 vs. Conor Daly and 2-0 vs. Ed Carpenter.) Tom Blomqvist mostly comes down to his 2014 European Formula 3 championship season, where he actually beat Max Verstappen in the championship and had a 17-7 record vs. Antonio Giovinazzi, 18-6 vs. Jordan King, 14-3 vs. Ed Jones, and 21-5 vs. Jake Dennis (I decided to include all the drivers expected to make a start in a major league racing series in 2021 even if they had not yet, although I did not count last weekend's Formula E races because I calculated most of this beforehand.) As you can see, most of the highest rated drivers that you do not expect are not hard to explain. However, one might argue some of these drivers are not true open wheel legends because their professional careers were not long enough or did not come close to matching their minor league performance. In several of those cases, I would agree.

One of the things I most like about this model is it does correctly identify major talents who are best-known for their successes in other disciplines: drivers who are recognized as great but mostly through their accomplishments in other series. In addition to the high placement of Supercars champion James Courtney, these include 3-time World Touring Car champion, José María López (.165), DTM champion Pascal Wehrlein (.112), and DTM championship contender Edoardo Mortara (.146.) Most of their significant accomplishments took place outside of open wheel series, but they were still great open wheel drivers in their own right. One thing shared by many of the top drivers on this list is indeed an eclecticism and willingness to cross over successfully between a bunch of different series, and that seems right.

Of the 449 eligible drivers who had 5 or more "teammate-races" on the initial iteration, Greg Moore was the highest rated at 0.425 while Taki Inoue was the lowest rated at -0.563. The average driver had a rating of -0.051 even though the model was designed so that 0 would be the average. This does make sense because there are usually at any given point in any given racing series more bad drivers than good. Additionally, the better drivers are more likely to have long careers, more likely to drive for multi-car teams, and more likely to drive for teams that have fewer mechanical DNFs, so the top drivers will be overrepresented in terms of the percentage of teammate relationships in which they participated. As a result, only 175 drivers (39.0%) were actually above average and the mean driver rating in the distribution was -0.051 with a median of -0.034.

However, not all racing series were created equal. The average overall driver rating for a driver who did have a teammate relationship in F1 was -0.000 (rounding to 0 but barely negative), while the average Formula E driver rated -0.003, the average IndyCar driver -0.050 (roughly in line with the overall distribution), and the average Super Formula driver -0.064. Generally speaking, this is how these series rate against each other by any metric. Formula 1 drivers are on average the best, Formula E drivers are not far behind, and IndyCar and especially Super Formula drivers are significantly behind. However, bear in mind that this period includes the IndyCar split so I will eventually recalculate this including post-2007 drivers only to account for this in a later paragraph.

When looking only at drivers who won a championship or won a race, the exact ranking appears when considering both of these metrics as well. The average driver rating for the ten Formula One champions is .132 (median .128), while the five Formula E champions average .116 (median .123), the 23 IndyCar champions average .077 (median .089), and the 17 Super Formula champions average .003 (median .002.) This seems to suggest a clear hierarchy in the prestige of each of these series since it remains consistent with the average of all the drivers in each series as well.

Although Formula 1, Formula E, and Super Formula had steady continuity throughout their history in this period (aside from the Japanese open wheel series changing its name several times from Japanese Formula 2 to Japanese Formula 3000 to Formula Nippon to Super Formula), IndyCar very much did not. The IndyCar split obviously very much complicates the evaluation of IndyCar drivers since there were at times two sanctioning bodies and at other times one, the depth of competition in both series could be wildly variant from year to year, not to mention this counts all oval races and road course races equally when most IndyCar drivers are better at one discipline than the other. Nonetheless, I do think a teammate-driven model is one of the best ways to account for this degree of chaos since the CART and IRL cars were usually similar enough that skill in one type of car is adaptable to skill in the other just as much as across any open wheel series. If Tony George had decided to revert to something akin to '60s roadsters for the IRL instead of formula cars, it would have likely been very different and the sprint car drivers he was catering his series to likely would have been significantly better, as opposed to what actually happened where it became a league for drivers who could not find steady rides in CART (some of them good, but many quite bad.) However, since both series used formula cars I see no qualms in throwing them all in together in one category just as IndyCar now itself does with their all-time win lists and even their YouTube channel (did anyone expect they'd actually upload the 1996 U.S. 500 onto it ever? It just happened...)

In order to compare how the current era of IndyCar stacks up to the other three open wheel series, I will instead look at only the drivers from 2008 to present and evaluate them. The average post-split champion had an average rating of .095 (median .081), placing it slightly closer to Formula E in terms of mean but further away in terms of median, indicating the typical post-split champion is not necessarily better than the typical champion during the split. However, since the post-split period had what most would argue was superior talent to the split period (with the probable exception of the 1996-2001 era of CART, which I think was probably deeper than the current era of IndyCar, 2000 and 2001 especially) it seems likely that the post-split drivers would tend to go up on future iterations due to the strength of their teammates while split-era drivers would tend down on average. Similarly, the median post-split winner had an average rating of .035 (median .040) while the median post-split starter had an average rating of -.050 (median -.034.) Again, these statistics don't really change significantly whether you select from the pool of all IndyCar drivers or just those who made starts after 2007. Perhaps the split did not have as much effect on the overall talent level as we thought?

First Iteration

Final Iteration

Generally speaking the drivers who gained the most when I reiterated the model where those drivers who had the strongest teammate average. This did naturally tend to benefit the drivers who competed in the stronger series and lower the ratings of the drivers who competed in weaker ones, although not always. While the absolute strongest drivers in general were in Formula One, not all of them were, and there were (especially on the weaker teams and especially in earlier years) lots of instances of teammates who were not especially strong. Generally, the early IRL and Super Formula drivers tended to fall, but not all of them (Scott Sharp rose a lot, Kenny Bräck rose slightly, and Norberto Fontana and Koudai Tsukakoshi gained considerably.) Regardless, the drivers who gained and lost the most should be a reasonably solid proxy for who had the strongest and weakest teammates.

The drivers who gained the most between the first iteration and the 30th were as follows: Alexander Wurz (who went from deeply negative at -.149 to .090 and gained a staggering .239), Fernando Alonso (.226), Michael Schumacher (.220), Lewis Hamilton (.215), Nico Rosberg (.192), Giancarlo Fisichella (.183), Valtteri Bottas (.176), Felipe Massa (.172), puzzlingly Thiago Medeiros (.172), Jenson Button (.167), Sam Hornish (.157), Adrian Sutil (.146), Ralf Schumacher (.145), Mark Taylor (.142), Sebastian Vettel (.141), Rubens Barrichello (.136), Kimi Räikkönen (.135), Gil de Ferran (.134), Gary Paffett (.132), and Simon Pagenaud (.128.) Wurz seems initially puzzling until you realize that his two F1 teammates were Fisichella and Rosberg, who both were the top ten most-improved drivers as the model reiterated. Most of the rest are the drivers who spent large portions of their career alongside championship-caliber teammates, which means most champions gained considerably as the model reiterated as you would expect. Anyone who was a teammate to an F1 champion rapidly improved on later iterations. Drivers outside of F1 did not do so much, but Sam Hornish and Gil de Ferran not surprisingly did because their major link was to Hélio Castroneves, who was clearly underrated on the original iteration. Once Castroneves improved, de Ferran and Hornish obviously improved even faster because they both beat him more often than not. Massa and Hornish at .146 and .144 were the two highest-rated drivers on the final iteration who were negative on the original model (because neither of them beat their teammates as much as Massa and Hornish's average teammates did, but when you consider Massa was teammates to three world champions and all of Castroneves's other teammates except for Ryan Briscoe were also IndyCar champions, this makes a great deal of sense.) Paffett mainly gains indirectly because Stoffel Vandoorne was his main link, and Vandoorne because he was Alonso's teammate just might be the overall most overrated driver in the entire model (ignoring things like the Mark Taylor anomaly.) Drivers who started out at or near the top generally continued to improve, so Taylor's overrating did not go away with future iterations: he just became more and more overrated.

Having said that, some drivers did go down, and some of them considerably. These included Dr. Jack Miller (-.401), Jüri Vips (-.339), Geoff Boss (-.279), Dan Ticktum (-.266), Yuki Tsunoda (-.265), Hiroki Otsu (-.255), Toshiki Oyu (-.249), Harrison Newey (-.248), Mick Schumacher (-.240), Tony Stewart (-.222), Laurent Redon (-.216), Ross Bentley (-.209), Ukyo Sasahara (-.204), Mark Smith (-.198), Mitsunori Takaboshi (-.196), Sarah Fisher (-.192), Jeff Ward (-.185), Willy T. Ribbs (-.183), Sena Sakaguchi (-.182), and a tie between Emilio de Villota and Jaroslaw Wierczuk (-.181.) You will notice almost entirely that these are drivers who competed during either the early IRL, late-period Champ Car, Super Formula, or mostly in minor league racing series. My decision to include all the drivers who have not made a start yet but are expected to like Tsunoda and Schumacher means those drivers are probably highly unreliable in the model since I am comparing drivers who were not yet major league drivers to drivers who had already made the major leagues (who were at that time better than them), so because they were losing in minor leagues to drivers who may have been more experienced, this suggests how unreliable the model is for inexperienced drivers (as I think does Mark Taylor and Rinus VeeKay on the other end.) I do think it becomes very, very accurate for drivers who have two or more full professional seasons.

Jack Miller is regarded by many as one of the worst IndyCar drivers of all time but his initial rating of -.105 was clearly hugely inflated because his three teammates in Indy Lights and IndyCar had absurdly high teammate winning percentages largely because they competed against him: Gualter Salles was 8-1 in general but 5-0 vs. Miller and Felipe Giaffone was 23-5 and J.J. Yeley was 1-0 but both of them beat Miller in their only races against him as well. Clearly, none of those drivers were remotely as good as their records as Salles despite an 8-1 record went negative and Giaffone's initial rating was nowhere near comparable to his record, although much to my surprise he did go up a rather solid amount. Since Miller dipped all the way below -.500, Yeley ended up having a score of exactly 0 in the model since I chose to not to penalize drivers for beating drivers rated below -.500. He was not quite the lowest-rated IndyCar driver though, as Laurent Redon, Dean Hall, Lyn St. James, Geoff Boss, Phil Giebler, and Luiz Garcia, Jr. all rated slightly lower, but none of them had very large sample sizes. Taki Inoue remained the lowest-rated F1 driver in the model at -.563 but Redon and Hall and St. James dipped below him by the final iteration, and Esteban Tuero made a sharp drop to -.559 narrowly nosing out Inoue. Since Inoue and Tuero are both often regarded as the worst F1 drivers of the last 30 years, this does indicate that the model is fairly accurate a lot of the time even for drivers with relatively few starts or shared races (F1 fans are wrong about Yuji Ide, though.) It seems like most of the inaccuracy comes from drivers who had significantly more minor-league starts than major-league starts (for instance Roberto Merhi's high rating because he had an exceptional junior record even though his F1 career was not impressive, or Daniil Kvyat rating over Carlos Sainz, Jr. because Kvyat had a better junior record.) Perhaps if I redo this model I really should throw minor league results out, but I feel I would lose too much data if I do that.

The big surprises to me are mainly how far Tony Stewart falls since he is the only driver on that list anyone would consider great in open wheel racing and also how far some of the drivers from CART's golden age (Ross Bentley, Mark Smith, and Willy T. Ribbs) drop. Stewart primarily drops because his main teammate link Robbie Buhl also dropped and Buhl dropped because his main links Stewart and Sarah Fisher both also dropped considerably. Buhl and Stewart sort of had a negative feedback loop going on that I think probably does underrate both of them in the same way that Mark Taylor, Richard Antinucci, and Ronnie Bremer had that bizarre positive feedback loop. Having said that, neither Stewart nor Buhl really could gain that much because they had a 3-3 tie in their teammate record (no matter how badly Stewart outperformed him in the actual races.) Since both started negative to begin with, they just dipped more and more negative. Stewart ended up by far the lowest-rated IndyCar champion (except for Buzz Calkins who did not have enough teammate-races to be eligible) and Buhl and Jaques Lazier ended up as two of the lowest winners, indicating that the speed of the Menard cars was to some degree masking their drivers' deficiencies. Do I think Stewart was that bad? No. But it is a chink in the armor for those who want to use Stewart's IndyCar champion as proof for his legend.

Despite competing in early '90s CART, Bentley was linked to a lot of marginal drivers including Buhl and losing to early IRL driver Alessandro Zampedri, so that makes sense. Mark Smith and Willy T. Ribbs despite also competing in that era seemed to have lowered each other much like Stewart and Buhl did since they were teammates. This likely also had a major effect on Scott Goodyear, who had an 11-0 record against Ribbs and a 10-0 record against Hiro Matsushita, which gave him a probably inflated rating at .009 at first, but he dropped to -.108 and then dragged all his teammates down with him. This interaction also indirectly affects Arie Luyendyk and Robby Gordon and is responsible for their declines, but it does make sense. Generally, all these drivers had on average exceptionally weak teammates even if they were competing in the glory years of CART. Goodyear lost to Arie Luyendyk 1-10 in two separate instances of being his teammates and as they both fell together, it seems inevitable that Goodyear would end up negative. I admit this may deserve a big asterisk since in 1990 when Goodyear lost to Luyendyk 1-8, he also had a slower engine because Luyendyk had the superior Chevrolet engine while Goodyear had the inferior Judd engine. Having said that, my attitude on these things is generally the best driver on a team will get the fastest cars for a reason (which is why I'm not all fussed about how team orders might affect this rating either) and also in this particular instance that Luyendyk also beat Goodyear 2-0 when they reunited as teammates in the IRL, indicating that maybe 1990 was a significant driver difference anyway; it's probably a little of both. And perhaps it's justified since he lost the championship to Buddy Lazier in the IRL in 2000 for Panther Racing before Sam Hornish utterly dominated him the next year in the same cars.

While a lot of drivers started negative and went positive because they had stronger teammates than average, there were actually surprisingly few who started positive and eventually went negative. These were: Toshiki Oyu (-.246), Jeff Ward (-.178), Chris Menninga (-.157), Sena Sakaguchi (-.153), Goodyear (-.108), Nicolas Kiesa (-.071), Shinsuke Shibahara (-.057), Oliver Askew (-.028), Álex Palou (-.023), Buddy Rice (-.011), Sho Tsuboi (-.007), Eliseo Salazar (-.006), Robby Gordon (-.005), and Eddie Cheever (-.003.) Note that these are almost entirely drivers from the split-era IRL and Super Formula, which had generally the weaker talent so this too makes sense. Kiesa is a big surprise since I did not expect to find an F1 driver on that list. However, Kiesa did get swept by his F1 teammate Jos Verstappen and was only positive to begin with because of his minor league links to late period Champ Car and IRL drivers Ryan Dalziel and Phil Giebler, respectively, who both went down. Askew is sort of puzzling because his only link was to Pato O'Ward and he went up. However, I think he mainly went down because O'Ward's initial rating was substantially worse than his record (because he benefited from being Dalton Kellett's teammate in Indy Lights.) Gordon, Salazar, and Cheever all got caught up as indirect collateral damage in that Goodyear/Smith/Ribbs triumvirate, especially Gordon because he was teammates to all three of them. I must admit I'm fairly surprised Salazar only dropped that low though. Despite being an Indy 500 winner, Rice generally had weak teammates: the only drivers he competed against who had more than one IndyCar win were Tomas Scheckter and Eddie Cheever, neither of whom would be considered all-time greats (although Scheckter did do unexpectedly well in the model.) Palou fell because his two main links, Santino Ferrucci and Tadasuke Makino, also fell significantly. Generally the drivers who gained and lost do make sense and it does seem to have quite a bit to do with the actual teammate strength. I would not have guessed Alexander Wurz of all drivers to have the highest average teammate strength though.

For the most part, the relationship between the drivers doesn't change a great deal. Fernando Alonso remains the highest rated Formula One driver across all iterations of the model. Mark Taylor actually jumped Greg Moore to be the highest-rated IndyCar driver but that was because he didn't actually have any teammate relationships in his one IndyCar season because he crashed so often; Moore remains the highest-rated driver among drivers who had actual teammate relationships in IndyCar. These three drivers were the only ones to rate above .5 so that the drivers who lost to them ended up losing nothing in the end. This probably especially inflated Stoffel Vandoorne, who lost to Alonso but none of his other teammates, and Patrick Carpentier, who lost to Moore but none of his other teammates, more than it should have, but maybe that is an indication that they are in fact genuinely underrated. I'll have to think about that more. The highest-rated Formula E regular through most iterations of the model was surprisingly Mitch Evans, not one of the champions (although he trailed Franck Montagny as well as part-timer Tom Blomqvist initially, who was wildly overrated because all his main accomplishments were in minor leagues; he was the highest-rated driver to see a slight decline) and Juichi Wakisaka for Super Formula (although Wakisaka trailed Peter Dumbreck and James Courtney, who had similarly short careers there; Wakisaka was clearly the highest-rated driver who made Super Formula his home despite not being a champion.) The model does have a general consistent order through most of the iterations, besides the corrections for the drivers who gained or lost simply because they had exceptionally strong or weak teammates, which indicates there does seem to be a lot of validity to this (not that I agree with every single ranking, mind you; some of them seem ridiculous to me but most do not.)

After reiterating all the drivers in the model, I next took averages of all the champions, winners, and starters for each series based on the reiterated model just as I had done before to determine how the four series stack up. The average Formula 1 champion has a rating of .282 (median .307), while Formula E champions average .191 (median .196), IndyCar post-split champions average .180 (median .187), IndyCar champions including the split years average .134 (median .171), and Super Formula champions average .028 (median .020.) Just as before, the four series rank in the same order talent-wise: Formula 1, followed by Formula E, IndyCar, and Super Formula. Formula E champions are substantially better than IndyCar champions on average, but since 2008 when Champ Car and IndyCar reunited there has basically been no difference (with a slight advantage to Formula E.) However, as expected neither really come close to Formula 1 champions, but they do come closer to Formula 1 champions than to Super Formula champions, many of whom actually have negative ratings. The relationship between the series does not really change even after many iterations. F1 is still much better than Formula E, which is barely better than IndyCar, which is substantially better than Super Formula. I'm not surprised Formula E is close and I'm not surprised it beats IndyCar since it is an international series and not a domestic one, its drivers were generally more successful in European feeder series, and there are fewer outright bad drivers there. I am surprised that Super Formula is further away from them talent-wise than they are to F1 though.

As far as comparing the series' winners to each other, the same patterns hold. Formula 1 winners average .200 (median .170), roughly comparable to Formula E and IndyCar champions, while Formula E winners average .123 (median .151), IndyCar post-split winners average .081 (median .103), winners including the split average .040 (median .030), and Super Formula winners average .014 (median -.013.) Formula E's advantage over IndyCar here is more substantial than it was in terms of champions, indicating that while the best drivers in both series are relatively comparable, Formula E clearly has more good drivers and IndyCar's weaker winners are substantially worse (it's hard to imagine Takuma Sato having as many wins in Formula E as he does in IndyCar; then again Felix Rosenqvist was more successful in Formula E than IndyCar to date so it's hard to say.) Interestingly, Super Formula's winners are not really far off from their champions as in the case with the other series, indicating that the championship itself is fairly random, which doesn't surprise me a lot since its season is significantly shorter.

Finally, I compared the starters of each series to each other just as before. Once again, I defined a "starter" as a driver who had at least one teammate relationship in that series, which excluded drivers who actually made starts but did not have a head-to-head result against a teammate, including one Super Formula champion (Ukyo Katayama, who became more famous as a mediocre F1 driver.) The average F1 starter has a rating of .054 (median .061), which is below the average of Formula E and unsplit IndyCar winners but above the average of Super Formula champions. In other words, most Super Formula champions do not belong in F1 except for the drivers who were just passing through who clearly already had interest in the F1 paddock beforehand like Ralf Schumacher and Pierre Gasly (the latter of whom didn't even win the championship.) However, it does seem based on this model that most Formula E and IndyCar winners are good enough to be F1 starters (indeed most of the Formula E winners to date already were either F1 starters or at the very least F2 championship contenders unable to find F1 rides, so any arguments that Formula 1 has the best 20 drivers in the world or even the best 20 open-wheel drivers in the world are completely unfounded. I do not think Scott Dixon or Mitch Evans would be F1 champions, but they'd be highly competitive.

Indeed the average Formula E starter has a rating of .045 (median .049), not far off of F1 starters at all. I suspect the actual difference is larger than this because the F1 average includes a lot of drivers from the '90s who are underrated by the model because for instance all of Alain Prost and Ayrton Senna's teammates did not benefit from Prost and Senna gaining in each reiteration like the teammates of Michael Schumacher, Fernando Alonso, and Lewis Hamilton did. This clearly has second-wave and third-wave effects throughout the model underrating most of the '90s drivers (I do not think Damon Hill, Jacques Villeneuve, Olivier Panis, and David Coulthard are the four worst F1 drivers to win a race as the model indicates, but simply that Hill and Villeneuve are significantly underrated, although I still think they'd be the two lowest-rated F1 champions regardless.) IndyCar and Super Formula have a lot more mediocrity though, and the back halves of their fields are substantially worse. IndyCar post-split starters were on average below average major league open wheel drivers (mean -.028, median -.014), and drivers during the split were worse still (mean -.047, median -.028.) Super Formula drivers as usual were the worst: their starters averaged -.075 (median -.070.)

So we have confirmation now: Formula 1 is substantially better than Formula E talent-wise, which is barely better than IndyCar (although fairly substantially better than the weaker IndyCar split fields), which is far better than Super Formula. This does make sense since F1 is considered the pinnacle, FE is international, while IndyCar and Super Formula are domestic. IndyCar's placement over Super Formula also makes sense because the US has a larger population and larger market than Japan does (Tora Takagi won a championship in Japan and was considered a joke in America, but to be fair, he was one of the weaker champions there too.) You can still probably make the case IndyCar is on the same level with Formula E though simply because the series requires more versatility given that it races on ovals, road courses, and street courses, and not merely street courses as in Formula E's case, but admittedly now that IndyCar only has three ovals left, its diversity is not the selling point it once was (you can even make the case that this year NASCAR is more diverse with its sudden focus on road courses and bringing back dirt to the Cup Series for the first time since 1970.) That doesn't mean that individual drivers from weaker series wouldn't be able to get strong results in stronger ones. A couple IRL split drivers (Scott Sharp and Kenny Bräck) do exceptionally well while there were quite a few bad F1 drivers (just fewer percentage-wise than the other series.) Just because a series is better doesn't mean a specific driver is, and there are all sorts of oddities where drivers overachieve in one series and underachieve in the other (the aforementioned Rosenqvist really should have done more in IndyCar based on his Formula E and Super Formula results, and I wonder if that isn't why Ganassi let him go.)

Before I display the complete results, there are three more minor things I'd like to mention. Even in my grad school statistics work, I have never done a robust iteration like this before and I haven't read much about how it should be done or how people did equivalent analyses in other sports. I knew I needed to keep reiterating until the data converged, but I wasn't entirely sure where to stop. I decided I was going to stop reiterating when all drivers' scores changed by a rating of less than .001 between iterations, but I ended up stopping a bit sooner than that because I noticed the scores were steadily creeping upward and there was an inflationary effect in the model. A lot more drivers' ratings improved that went down. The average driver on the first iteration had a rating of -.051, while the average driver on the 30th and final iteration had a rating of -.042. That isn't a huge difference by any means but on the last few iterations, the vast majority of drivers (even those well in the negatives) were gaining and few drivers were losing. I don't know what caused this, but I think it might have been the fact that I set all ratings above .5 to equal .5 and all ratings below -.5 to equal -.5 when I reiterated so drivers did not lose anything by losing to a great driver and drivers did not gain anything by beating a terrible one. I stopped at 30 iterations because I didn't like the inflation effect, but looking at the drivers who came out around average, it does seem right as those are the sorts of drivers I do pretty much consider to be average myself (with the possible exception of Charlie Kimball and Carlos Huertas, who seem a bit inflated to me, but maybe not - Kimball was after all usually going up against championship-caliber teammates in IndyCar and Huertas did win a race as a rookie in a Dale Coyne car), so I think it turned out pretty well.

Additionally, you will notice that there were some rounding errors on this below table that indicates how each driver shifted from the 1st to the 30th iteration. For instance, Michael Schumacher went from .182 to .403 but the change in his rating is listed at only .220, not .221; this is simply due to rounding. I did notice while I was finishing up my calculations that there were several Formula E driver records I had slightly off because unfortunately Wikipedia does not always do a good job marking DNFs on its seasonal finish tables (very often a driver did not finish a race but it is not marked in purple as is the Wikipedia racing table convention.) I notice this mistake a lot and had to check and double-check things very often, but I failed to do so in this case until this column was about 80% written so I decided not to recalculate anything after that. Regardless, these are all minor errors that probably wouldn't have affected any scores very much if at all.

Needless to say, I don't personally agree with all these ratings and some I find ridiculous but in most cases I do understand and largely agree with them. Even the ratings that seem off-the-wall usually make sense when you look at the actual teammate relationships. For this list, I would say that a driver needs two or more major-league seasons with teammates to likely ensure an accurate rating, but I do think a large percentage if not most of the ratings for drivers who do not have this are still fairly accurate. If you asked me to evaluate the driver ratings and categorize certain ranges for them, I would say that in general, drivers with a rating >= .2 are career-long elite, drivers >= .1 and < .2 are great, drivers >= .05 and < .1 are very good but not great, drivers >= 0 and < .05 are good, drivers >= -.05 and < 0 are mediocre leaning good, drivers >= -.1 and < -.05 are mediocre leaning bad, drivers >= -.2 and < -.1 are bad but not awful, and drivers < -.2 are very bad (usually these are the drivers people call ride-buyers.) That's not to say there aren't outliers in every category (I certainly wouldn't call Tony Stewart bad or Mark Taylor the best driver of this era), but in general, the drivers I consider great are all rated above .1 and the drivers I do not are rated below it, and this separates out rather nicely.

Formula 1 champions ranked

Most statistical models these days based on teammate-to-teammate relationships seem to agree that Fernando Alonso was the greatest driver in recent Formula 1 history. He beat every teammate he ever competed against and he never had less than a .500 record against any teammate in his F1 career in even an individual year, which is almost unheard of (Alonso did tie Jarno Trulli 4-4 in 2004 and Jenson Button 4-4 in 2015, but all his other years were at least 1.5-1 blowouts regardless of teammate: even though Lewis Hamilton beat Alonso in his 2007 rookie season when Alonso was the two-time defending champion, Alonso actually beat him 9-6 in the races that year.) Michael Schumacher would probably be close if you threw out his 2010-12 return, but I don't think you can do that. As far as the comparison between Alonso and Hamilton, in addition to beating Hamilton straight up in 2007, Alonso usually did better in their relationships with linked teammates. One exception is that Hamilton did beat Jenson Button (24-13) slightly worse than Fernando Alonso did (11-8), but I suspect the unreliability of the McLaren cars when Alonso was there the second time hurt his record there. More tellingly is Alonso and Hamilton's relative performance against Felipe Massa and Valtteri Bottas respectively. When they were teammates, Massa and Bottas were very evenly matched with Massa beating Bottas 22-21. However, Alonso beat Massa (58-6) by a far greater margin than Hamilton beat Bottas (51-21.) It is true that Ferrari enforced team orders more than Mercedes did and Bottas was probably allowed to compete with Hamilton much more than Massa was allowed to compete with Alonso (Massa was ordered to pull over for him after all, which is almost offensive in retrospect when you consider that Nelson Piquet, Jr. crashing for Alonso in 2008 cost Massa that championship.) Having said that, I don't think the team orders difference is remotely large enough to make up for the huge discrepancy of a 91% winning percentage versus a 71% winning percentage. Schumacher would probably be ahead of Hamilton if you neglected his Nico Rosberg years, but since the model doesn't, it makes sense that Hamilton (who beat Rosberg) would be ahead of Schumacher, who lost to him. Vettel, Rosberg, and Button's rankings are pretty sensible too. It makes sense that they're all not very far apart from each other, but they should probably be ranked in that order (Rosberg at 27-37 definitely came closer to Hamilton than Button at 13-24, in addition to winning a title over him, which Button did not.)

The only real flaw I see when ranking the F1 champions is that I believe the '90s drivers are significantly underrated due to the vagaries of the model (and this also applies to the IndyCar and Super Formula drivers of that decade.) Although I included a lot of '90s drivers as well as post-2000 drivers who competed in the '90s, since some of those drivers' teammates aren't listed in the model, their ratings never re-iterate. For example, Damon Hill lost to Alain Prost 5-4 in 1993, which was probably one of his best ever seasons. However, Prost is not yet included in the model and only had a career teammate record of 46-31 (probably lower than it should be because he tended to have exceptionally strong teammates.) If Prost were included in the model, he like most legends would probably accrue higher and higher ratings with each future iteration and his actual driver rating would be far above his lead change record of 46-31, which at .597 would correspond to a .097 rating here. Since his rating is likely well above .3 if not above .4, his three teammates who are eligible for this list (Eddie Cheever, Jean Alesi, and Damon Hill) are all underrated, and therefore all their teammates are underrated, and so on. I will never be able to fix this until I've literally entered everyone in open wheel history, and that would still take several more months. As a result, I do expect that either directly or indirectly all '90s drivers are underrated and Hill and Villeneuve (along with elsewhere Al Unser, Jr. and Michael Andretti) should be higher than this. Don't get me wrong: Hill and Villeneuve would not gain enough to not be the two lowest champions in the model - they're too far behind anyone else for that not to be the case. But I don't think when I have calculated this in full that they will still be below .1 or anything like that. You can tell how biased the model is against '90s drivers by simply identifying the lowest-rated F1 winners in the model, which are David Coulthard (the only negative-rated winner), Villeneuve, Hill, and Olivier Panis, who are rated well below even Pastor Maldonado for instance. Would anyone really say they're the four worst winners? No, probably not. So throughout this model you should probably take a lot of the '90s drivers' ratings with a grain of salt. Not so much Michael Schumacher or his teammates though: Schumacher's main teammates Rubens Barrichello, Johnny Herbert, and Eddie Irvine were all teammates to each other in addition to being teammates to Schumacher so they have relatively few links to exclusively '90s drivers... I don't think those drivers will gain nearly as much as the drivers who have important connections to Prost, Ayrton Senna, or Nigel Mansell for example.

IndyCar champions & Indy 500 winners ranked

The top drivers on this list are not very surprising. Scott Dixon leads nearly every major IndyCar statistic among drivers of his generation except Indy 500 wins and has beaten all but two teammates in his career, so his leading by a fairly large margin is not surprising. I was mildly surprised Gil de Ferran nosed out Josef Newgarden and Will Power for second place, but I wasn't that surprised since he did peak in 2001, which I earlier calculated as the most competitive season in IndyCar history and he also retired before he truly declined. Despite this ranking, I am more impressed with Newgarden than any other 21st century IndyCar driver other than Dixon considering he has now won the most races in four consecutive seasons (becoming the first driver to do that against an unsplit field since A.J. Foyt in the '60s) and he at one point had the highest lead change since the '50s. Dixon and Newgarden are both doing things that no other IndyCar driver has done in decades. I feel Power is a little overrated here because he crashes a lot and I don't count DNFs against drivers in this model, but he is only narrowly behind Newgarden in his teammate head-to-head so that is reflected in this result. I suspect it would be much wider between Newgarden and Power if I counted Power's crashing against him.

The major revelation here is the astonishingly high ranking of Scott Sharp, the most misunderstood driver of the CART/IRL split period. CART fans sneered at him because he crashed in the first corner after winning the pole at the 2001 Indy 500, a race where the CART teams would lap the field. He wasn't really the IRL fans' favorite either because he came from road racing and never seriously contended for an Indy 500 win. He was often not seen as a "real champion" because he failed to win a race in the three-race 1996 season and only shared in the championship because there wasn't a tiebreaker, but he was actually surprisingly ranked better than a lot of the more widely accepted champions. Sharp was reliably and consistently dominating teammates for an extremely long period: from his CART rookie season of 1994 to his last full-time IRL season in 2007, he beat every single teammate he ever competed against from one-offs to full-time teammates. Even Dixon, Newgarden, and Power did not beat every teammate, although Newgarden only lost to part-time teammate Juan Pablo Montoya and de Ferran did also beat every single IndyCar teammate. Sharp did not merely beat his teammates: he beat every single teammate in this period by at least a 2-1 margin, except Al Unser, Jr. who he beat 11-8. Although Unser was certainly not at his peak given that his arrest occurred during this period, Sharp is still the only driver to have a head-to-head winning teammate record against Unser in his entire career. Bobby Rahal didn't do it. Danny Sullivan didn't do it. Emerson Fittipaldi didn't do it. Paul Tracy didn't do it. Sharp did. Although most of Sharp's teammates were not that great, he did beat two other prolific winners he competed against part-time: Adrián Fernández and Ryan Hunter-Reay in addition to the fairly obscure driver Jeff Simmons, who is also surprisingly highly rated in the model. The only two teammates Sharp ever lost to were in part-time one-offs that bookended his career: against Stefan Johansson in his 1993 debut and against Dan Wheldon in the 2009 Indy 500, and there's nothing embarrassing about either of those.

Sharp's sneakily legendary teammate record has caused me to rethink quite a few things about how I evaluate drivers. For instance, on my 100 greatest IndyCar drivers list, I never even considered him for it despite his 9 wins and his admittedly farcical championship when it now seems like he should have been an obvious lock. When I was working on that, I decided to weight each driver's performance based on the era in which they competed, which was not good for Sharp because the early IRL was exceptionally weak. But this has taught me something. Just because a driver is amongst a weak field does not mean that driver is weak. A driver's record may be inflated due to a lack of competition, but that doesn't mean the level of talent changes. And there were many, many pieces of evidence in Sharp's favor: his 2 Trans-Am titles before his career and his ALMS title afterward (very few of the other early IRL drivers actually had won major league championships elsewhere), his two 24 Hours of Daytona wins 20 years apart, the fact that he was the only one of the original IRL drivers who was still able to win after the CART drivers and teams crossed over, the fact that he was the first driver since A.J. Foyt himself over a decade earlier to win both a race and a title in his cars, the fact that he was the first IRL driver to win 7 seasons in a row, and even his only NASCAR Winston Cup race when he gave Jimmy Means his last lead lap finish as a car owner. Sharp was really something of a legend, and I was too blind to see it because he was competing against many bad drivers.

On the flip side, the big shocker in the opposite direction is how far Sharp's onetime rival Tony Stewart is behind all the other drivers on this list: he's not even close to anyone, even the four drivers usually considered the worst Indy 500 winners of this era (Takuma Sato, Buddy Rice, Eddie Cheever, and Buddy Lazier.) While their ratings aren't that surprising, Stewart's definitely is because he was the IRL's poster boy and considered the dominant driver on the scene (he did after all lead the most laps in seven consecutive races in 1997.) However, it now seems clear that Stewart's dominance was more due to the fact that Team Menard's horsepower had a massive advantage over all other teams that masked its drivers' deficiencies. While I certainly don't believe Stewart was that bad, he did have a learning curve going from front-engine sprint cars to rear-engine formula cars just as all other sprint car drivers did, which is why in recent decades those drivers have taken years to get up to speed in IndyCar racing. Billy Boat and Davey Hamilton actually did a little better at -.114 and -.186 but they also had experience in formula cars before 1996 between Boat's Indy Lights season and Hamilton's previous 500 attempts while Stewart entered green in 1996 and was also distracted by his NASCAR Busch Series forays. Stewart had an extremely small sample size because he didn't have teammates for many of his best races, but his 3-3 record vs. Robbie Buhl was not very good, and then he finished last of the three Ganassi drivers in the 2001 Indy 500 giving him a career record of 3-5 and since Jimmy Vasser was just barely above average that cost him a lot too. He certainly ran much better than this finishing record implies but even if you account for that he still might be last. The Menard cars were just that dominant. If I counted Greg Ray and Buzz Calkins, who had even smaller sample sizes, Ray and Stewart would have been 2 of the bottom 3 champions in the model (both ahead of Calkins.) And Stewart was the lowest-rated IndyCar winner with 5 or more "teammate-races" in the entire model with Jaques Lazier and Robbie Buhl 2nd and 3rd, and both of them won in Menard cars too. That's how dominant they were. The early IRL was often hyped as a rivalry between the Menard team and the Foyt team, which were considered equally strong. The reality was Foyt only appeared as strong because they had Scott Sharp and Kenny Bräck driving for them. For all he is mythologized as Gen X's equivalent to A.J. Foyt and Mario Andretti in terms of versatility, he does not live up to that hype. Foyt and Andretti were much more better overall drivers than Stewart were, although I will grant that Stewart was a much better NASCAR driver than they were.

After Sharp, a large gaggle of drivers are nearly tied between ratings of .17 and .18. I think Dan Wheldon, Cristiano da Matta, and Kenny Bräck are all a little overrated here because they all seemed to be in decline for a year or two before their career-ending crashes. If their careers were as long as they should have been, their late-career years would have sent them plummeting like it did for Juan Pablo Montoya, who I suspect would be over .2 on this list if he had not returned to IndyCar in 2014, where he became Hélio Castroneves's only full-time teammate to lose to him (even Ryan Briscoe actually beat Castroneves.) On the flip side, I think Kanaan, Bourdais, and especially Montoya are all underrated because they did have entire careers including ups and downs that Wheldon, da Matta, and Bräck didn't really have to the same degree. I believe Michael Andretti and Al Unser, Jr. (and considerably further down the list Jacques Villeneuve) are especially underrated because of the bias towards '90s drivers.

Quickly summarizing the rest of this list, I am not surprised to see Dario Franchitti and Paul Tracy basically tied since they were considered evenly matched discounting Tracy's crashing (which this model does discount), nor am I surprised to see Franchitti behind drivers with seemingly weaker stats like Tony Kanaan and Dan Wheldon because they utterly dominated him. Franchitti, like Montoya, seemed to lack year-to-year consistency as if when Franchitti wasn't in the championship hunt, he didn't try as hard, thus making him very championship-or-bust and Montoya seemed the same way on his IndyCar return. At their peaks, both of them likely would be top five drivers on this list, but there were too many races and even seasons when they weren't at their peak. Alexander Rossi was way lower than I expected: according to popular opinion, he and Josef Newgarden are basically tied for 2nd behind Scott Dixon, but I always thought Newgarden was clearly better. Definitely not to the degree indicated by this model though. It makes me wonder: is this another Stewart situation where Andretti actually has the fastest cars but doesn't have the right drivers occupying them? They've certainly had the fastest cars more often than not at Indy in the last decade. Because Ryan Hunter-Reay was obviously even lower and I noticed that the Andretti IndyCar drivers from the 2010s all stayed about the same or went down in future iterations while the Penske and Ganassi drivers tended to rise substantially, it raises a question. Rossi is definitely closer to the tier of drivers who may or may not win championships depending on how lucky they get than to the kind of drivers who win dominant championships and having clearly been overtaken by Colton Herta, this seems to provide evidence that Rossi might be less likely to win one than everyone seems to think. A large number of unlikely things had to happen at once for Hunter-Reay to stumble into a championship, and Rossi is closer to him than he is to the true legends. I've talked about Alex Zanardi's overratedness before and that was no surprise to me, but it fits the data: this is about where he belongs having beaten Jimmy Vasser badly but also been beaten badly himself by Kanaan. I do find it interesting to see the 1996 Indy 500 champion nosing out the U.S. 500 champion in the wake of the split, but I can't say it surprises me. Buddy Lazier does have a very small sample size since his Hemelgarn Racing team was usually a single-car operation, but he did pass Vasser for 2nd late in the 2000 Indy 500 and passed Franchitti and Kanaan late in the 2005 race and he clearly had inferior cars both times considering Vasser, Franchitti, and Kanaan's teammates (respectively Montoya and Wheldon) won those races. And he did win the Indy 500 with a broken back passing a guy who won the 24 Hours of Le Mans less than a month later rather than causing a multi-car crash on the pace lap and winning in his backup car. (I kid, I kid. I like both of them.) For all the vitriol CART sent the IRL's way, in retrospect a handful of the IRL's biggest stars were still better than most of the CART drivers (your mileage may vary on Stewart.) Having said that, Lazier was no Sharp either.

Formula E champions ranked

This will not be nearly as complicated as analyzing the IndyCar drivers as there was no split in this series and the series has only existed for six seasons so there have only been five champions. The big surprise here is that the series's all-time win leader Sébastien Buemi is by far and away the lowest rated of the five, but looking into it more deeply, this isn't as surprising as it seems. Buemi, like Tony Stewart (but with a much larger sample size) was extremely dependent on having the fastest cars on the track: his e.dams team won the manufacturer's championship in Formula E the first three seasons but he only won one of those titles, which means that the two champions in those years (Nelson Piquet, Jr. and Lucas di Grassi) likely had worse equipment, which explains their higher ratings. When Buemi stopped having the dominant cars, he has mostly since stopped winning and has even tied his teammate Oliver Rowland with one win each. It's pretty telling that Buemi's teammate in his heyday Nicolas Prost had a rating of -.178 despite winning 3 races for the team, which implies that the cars were so dominant that a lot of drivers could have won in them. Additionally, when he was in Formula One, Buemi was tied 12-12 with his teammate Jaime Alguersuari, who was a good driver but not particularly elite, while di Grassi, Vergne, and Piquet all arguably had stronger F1 records. Buemi in fact himself lost to di Grassi 1-7 in GP2 in 2007, which may explain why he is so far behind here. I do think Piquet may be slightly overrated because his teammate was Alonso and Alonso rose so highly that even losing teammate matchups to him cost him nothing. Perhaps it's not surprising that António Félix da Costa leads the way because unlike the other four drivers, he did not have a Formula One career at all and therefore he has been dominant for his entire professional career while the others have not. Although da Costa's minor league record was also arguably the most impressive: although he did get blown out 17-1 by Valtteri Bottas, he has a winning record against all 22 of his other teammates in his entire career, including Kevin Magnussen (21-3!), Alexander Rossi (8-2!), Josef Newgarden (2-0), Carlos Sainz, Jr. (2-0), Vergne himself (5-4), as well as his undefeated teammate record in four different Macau Grands Prix. If da Costa beat Vergne and di Grassi beat Buemi, this ranking is pretty straightforward. Nonetheless, all of these are great drivers and da Costa clearly seems to be one of the best in the world. If Formula E drivers are in fact reasonably close to Formula 1 drivers, da Costa did win last year's championship by almost 2-1, and that is incredibly astonishing. Many people criticized da Costa's third-place ranking in Autosport's world driver rankings for 2020 but that seems to be justified for me. There wasn't any more obvious selection for 3rd place after Lewis Hamilton and Max Verstappen (whichever way you decided to rank them, and for 2020 I'd have preferred Verstappen.) I think this series gets a little underrated because it doesn't get as good TV coverage and because of the gimmickry of the FanBoost and because people don't like not being able to hear the engine sounds; the competition is clearly excellent.

Super Formula champions ranked

This series is noted for its mix of Japanese native drivers and international drivers from elsewhere who tend to be better known for participation in Formula One or sports car racing; since many of the Super Formula drivers also simultaneously race in Japan's Super GT sports car series, these drivers tend to be very versatile. Super Formula can also be a stepping stone for drivers who eventually succeed elsewhere, such as Felix Rosenqvist and Pato O'Ward in IndyCar or Stoffel Vandoorne in Formula E or Pierre Gasly in Formula 1. However, the depth of competition does not match that of Formula E or IndyCar, much less Formula 1. Having said that, there are definitely still great drivers here, but as you can see, usually the best drivers are those from outside Japan. The highest-rated driver João Paulo de Oliveira is not that well-known outside of Japan, especially compared to F1 star Ralf Schumacher and Le Mans winner André Lotterer right behind him, but he has been a consistent force in both Super GT and Super Formula consistently since 2007, winning a race every year from 2007 to 2016 in one or the other series despite not seeming to have the fastest cars. He beat the much more famous eventual Super Formula champion Kazuki Nakajima 13-3 when they were in Japanese Formula 3 together before going on to beat every single teammate in his Super Formula career through 2016 aside from a one-off loss to James Rossiter in 2018. This included an 8-5 record against fellow champion Tsugio Matsuda, who was actually by far the lowest-rated champion in the model and a 5-0 sweep of ex-Formula One driver Narain Karthikeyan. de Oliveira also made a one-off start at the IndyCar Motegi road course race in 2011 where he qualified in the top 12 in a fairly slow Conquest Racing car and remained there until a mechanical failure. Although there were actually non-champions who ranked higher, de Oliveira is probably one of the greatest drivers you might not have heard of.

As I mentioned, the most striking thing about this list is the difference between the Japanese and the non-Japanese drivers. Among the nine champions rated above average, two are Japanese and seven are international. Among the nine champions rated below average, six are Japanese and three are foreign. The average non-Japanese champion had a rating of .080 to -.037 for the Japanese drivers. A lot of drivers tried to use this series the way many college basketball players go one-and-done into the NBA, most successfully when Ralf Schumacher won the title in his rookie season in 1996 and never competed in it again, using that as a launching pad for his successful F1 career, but sometimes it didn't go as well: Ralph Firman also immediately went to F1 after winning the Super Formula title (then called Formula Nippon) and Tora Takagi followed a dominant 2000 Formula Nippon campaign with a disappointing IndyCar career. It seems from what I can tell that the Japanese drivers are more likely to get access to the faster cars, leaving the Europeans and South Americans to overachieve in slower cars. The only two Japanese drivers who were rated above average both had many championships: Satoshi Motoyama won four championships in 1998, 2001, 2003, and 2005, while Naoki Yamamoto won three in 2013, 2018, and 2020. The three other multi-time Japanese champions were all rated below average with Kazuki Nakajima both at Hiroaki Ishiura at -.016 and Matsuda rating at -.239 well below anyone else despite winning two titles. Unlike Tony Stewart, this was not in part a case of small sample sizes as Matsuda had a career teammate record of 17-46; even in his winless 2007 championship season, his teammate Benoît Tréluyer, who won the previous year's title, beat him 5-1 but Matsuda beat Treluyer by 1 point because Treluyer had 3 DNFs and Matsuda had none. Clearly, both benefited from their Team Impul being utterly dominant as Tréluyer was also one of the lowest-rated drivers in the model. Although Tréluyer is very well-known for his sports car success, his Audi teammate Lotterer was always considered the team leader and this ranking makes it clear why. Another thing I notice is that the more recent champions are substantially lower-rated than the earlier ones for the most part. The teammates Hiroaki Ishiura and Yuji Kunimoto, who like Tréluyer and Matsuda won three straight championships for their INGING team clearly also benefited from very dominant equipment if both of them rated negative. Indeed, every champion from 2012 to present except Naoki Yamamoto was related below average by the model, perhaps explaining why Yamamoto could be a three-time championship winner and only rated at .053. It seems the glory days for this series in terms of competition may have been the years from 1996-2012 when it was named Formula Nippon; in the years since it became Super Formula the competition has seemingly declined, especially after André Lotterer left the series for Formula E.

Highest-rated drivers to never win a Formula 1 championship (2 or more seasons)

The top three names on this list should come as no surprise to anyone. Max Verstappen has been Lewis Hamilton's closest rival over the past couple years even though he hasn't had a fast enough car to contend for the championship (yet.) Charles Leclerc took over team leadership at Ferrari over Sebastian Vettel in only in his second year. Daniel Ricciardo was along with Leclerc the only teammate to beat Vettel in his career. Those three are by a large margin the highest-rated drivers to never win the championship. Verstappen is rated higher than all but three champions (Fernando Alonso, Lewis Hamilton, and Michael Schumacher.) Leclerc and Ricciardo were around the median for champions as they rated below five champions in the model (those three, Vettel, and Nico Rosberg) but above the other five. Any of these three probably could win/have won a championship in the right car and I imagine Verstappen and Leclerc probably will, but I think time is likely running out for Ricciardo. I think Ricciardo and Leclerc may both be a little overrated because of the weights there 2014 and 2020 respectively seem to carry here. I have read that Red Bull supposedly gave Ricciardo preferential treatment over Vettel in 2014 because they knew he was leaving, and the same thing might have happened with Leclerc last season. If true, that likely means this model overrates both of them because it does not take that into account. Stoffel Vandoorne is likely overrated because his F1 teammate was Alonso and since Alonso had a rating over .5, he lost nothing by losing to him (which is probably the same reason Nelson Piquet, Jr. is overrated as well.) Having said that, I do think Vandoorne is pretty great because he dominated the 2015 GP2 season by a huge margin over Alexander Rossi and became the first driver to ever win races in both Super Formula and Formula E.

Timo Glock is best known as either "the guy who en route to his first Champ Car win was forced to pull over handing Oriol Servià his first win", a GP2 champion, or "the guy who Lewis Hamilton passed on the last lap to win the 2008 World Championship," depending on your perspective, but he deserves a much greater legacy than that. In his one Champ Car season as a rookie, he beat third-year driver Ryan Hunter-Reay so badly that RHR was fired and unable to find a ride in either Champ Car or the IRL for nearly two years. Considering Hunter-Reay had won a race in his first two seasons in Champ Car, for Glock to render a future Indy 500 winner and champion that irrelevant that he became rideless for years is very impressive. But he did not stop there. In F1, he had four full-time teammates and beat all of them and they were all good to great and very highly rated by the model: Jarno Trulli, widely and correctly regarded as one of the most underrated drivers of this era (he too appears on this list), Formula E champion Lucas di Grassi, Formula E winner Jérome d'Ambrôsio, and even his least well-known teammate Charles Pic was still rated around 2010s legend André Lotterer by the model. He did lose to Nick Heidfeld part-time in 2004, but there's nothing remotely disappointing about that as Heidfeld beat several other very good to great drivers in his career in addition to Glock (including Kimi Raikkönen, Heinz-Harald Frentzen, Mark Webber, and Jacques Villeneuve.) Glock was a truly great open wheel driver - the last IndyCar graduate who arguably succeeded in F1 and it's about time he was recognized for it.

Valtteri Bottas had a rather low rating on the initial iteration but gained substantially because his two main teammates Lewis Hamilton and Felipe Massa also did. Mika Salo, like Trulli, is well-known for his underratedness and he would have been a winner had he not been ordered to play a support role for Ferrari in 1999 while substituting for the injured Michael Schumacher. Paul di Resta is likely overrated because his high rating largely stems from his championship in the 2006 Formula 3 Euro Series, where he beat among others his teammate Sebastian Vettel. di Resta did not find an F1 ride immediately when he came of age, so like many near-miss F1 drivers he ended up in Germany's DTM touring car series. By the time he made it to F1, his moment had passed and he wasn't a championship-caliber driver. Judging by the fact that drivers in the .2 range like Valtteri Bottas and Giancarlo Fisichella were utterly dominated by their teammates Lewis Hamilton and Fernando Alonso, it seems that a .2 rating is not good enough to seriously contend for F1 championships today, although it seems like it used to be in the past, indicating that the driving standards in F1 have risen. Verstappen and Leclerc are probably the only drivers on this list who'll deserve a championship at some point in their careers.

Highest-rated drivers to never win a Formula 1 race (2 or more seasons)

The top five drivers have been sufficiently discussed elsewhere, but I have a few comments on the others. I honestly expected Nick Heidfeld to be overwhelmingly the highest-rated non-winner since in my opinion he was the best, but that didn't turn out to be the case. It is true that when he switched to Formula E after his Formula 1 career ended, he also went winless there while two of the drivers rated ahead of him (Stoffel Vandoorne and Jean-Éric Vergne) did not. I think Heidfeld is probably slightly hurt relative to some of these other drivers because his chief link was to Robert Kubica and Heidfeld had a slight losing record to him (20-22) and he was therefore hurt immensely by Kubica's comeback when he posted a 2-16 record against his rookie teammate George Russell. That dropped Kubica quite a bit behind Heidfeld at .156 but I suspect if you exclude that season (when Kubica returned to F1 after an injury and a long hiatus and probably wasn't really the same driver as he was before), both of them would gain a lot and Kubica might even jump Heidfeld. Having said that, Heidfeld still has an excellent rating considering the number of good-to-great teammates he either tied or beat. Nico Hülkenberg is also widely cited as one of the best drivers in recent years to not win a race, even though he never ended up sitting on a podium. His career definitely is undeniably impressive regardless though as he posted a winning record against Paul di Resta, tied Sergio Perez, and beat Carlos Sainz, Jr. 2-1. The only full-time teammates he lost to were Rubens Barrichello in 2010 and Daniel Ricciardo in 2019.

The other three drivers might require a bit of explanation since they are not nearly as obvious. Charles Pic kind of puzzles me because generally speaking throughout his career he beat all his bad teammates and lost to all his good teammates, but I think he largely benefits from the high rating of Timo Glock, who he only lost to by a margin of 5-7. Pic did better against Glock than Lucas di Grassi and not much worse than Jarno Trulli did against him (11-12), so from that perspective it's kind of understandable. Pic's other highlight was a very impressive season in the Formula Renault 3.5 Series in 2009 when he beat Brendon Hartley 10-1 and also beat Daniel Ricciardo in a one-off start; I think that season may be doing a lot of the heavy lifting here. Romain Grosjean's most impressive results seem to have come in the junior series, particularly in 2007, where he beat Hükenberg 10-7 and Kamui Kobayashi 14-2. His F1 career was by no means bad as he was only blown out by two world champions (Fernando Alonso and Kimi Raikkönen) but he did also barely lose teammate matchups to Esteban Gutiérrez, who weren't ranked all that highly in the model so I think his junior results are doing most of the work here. Adrian Sutil is maybe the most surprising of all, but I guess he primarily gained from close losses to Paul di Resta (13-14) and Giancarlo Fisichella (5-7), who were significantly higher rated. I think di Resta is likely overrated because he had a much larger gap between his junior career and his actual F1 results than most drivers and that also made Sutil overrated.

Highest-rated drivers to never win a Formula E championship (2 or more seasons)

While it is no surprise that Fernando Alonso was the highest-rated Formula 1 driver, the big surprise is that none of the highest-rated drivers in the other three series ever won a championship: Greg Moore in IndyCar, Juichi Wakisaka in Super Formula, and arguably most surprisingly Mitch Evans in Formula E, who nosed out Scott Dixon to become the highest rated New Zealand driver in the model. While many people think Moore would have eventually won either a CART or IRL championship for Penske in the 2000s if he had lived and Wakisaka although unknown outside of Japan is a major legend domestically with three Super Formula championships, Evans relative to the other two sort of flies under the radar and isn't as noticeable as first, but he might just be both the most underrated driver in the world and the best driver on today's racing scene you've never heard of. While Evans is not very famous nor were most of his teammates over the years, his career teammate record of 79-19 is pretty astonishing (among all drivers with 50 or more "teammate-races", only Franck Montagny and Fernando Alonso had higher winning percentages against their teammates.)

Evans's Formula E teammate record of 23-9 isn't as dominant, but not a whole lot of drivers beat their teammates overall by a 2.5-1 ratio. He had a winning record against every teammate he competed against from the start of his career to present, and he beat three major league champions, beating future Super Formula champion Nick Cassidy 16-2, A1GP champion Adam Carroll 6-3 (A1GP was considered a major league when it was running, but now no longer is), and Formula E champion Nelson Piquet, Jr. 6-4. However, that underestimates how badly he beat Piquet. After six races in the 2018-19 season, Evans finished 9th or better in all of them and scored 36 points while Piquet's best finish was a single 10th place finish and he scored one point, after which he left the Jaguar team and would never reemerge in the series again. To dominate a past champion teammate - indeed the first champion and possibly its most famous driver due to being the son of a past F1 champion along with his own F1 and NASCAR experience, not to mention the Singapore controversy - so badly that it ended his career says a lot for Evans. The race after Piquet left, Evans immediately scored his first win and backed that up with another win the next year. He did this despite rather slow cars: both seasons Evans's Jaguar team finished 7th in the manufacturer's championship; probably very few drivers could have won in those cars at all when you consider that even Sébastien Buemi, the series's all-time winner can't win very often when he doesn't have the fastest cars. Considering all this, I don't really find Evans's ranking all that spurious. In a sneaky way, he does really seem to be one of the best drivers in the world.

This list along with all the others is reflecting how drivers did across all series and not merely the series in question, so some drivers here are overrated because of their successes in other series: Nick Heidfeld and Felipe Massa were clearly more successful in Formula 1 than in Formula E because they were past their primes when they moved into this series, and the same likely applies to Stoffel Vandoorne whose junior record and Super Formula season were probably more impressive than his Formula E seasons (although he has definitely been very impressive in this series as well), and André Lotterer seems to have declined from his peak from around ten years ago. Edoardo Mortara and especially José María López are better known for their performances in touring cars than their performances in this series particularly with Mortara finishing 2nd in the 2016 DTM championship and López winning three straight World Touring Car Championships from 2014-16 but although they like Evans didn't really have fast enough cars to win often, their talent was still definitely felt relative to their teammates. Mortara beat Felipe Massa 11-3 across the 2018-19 and 2019-20 seasons, tying of all people Michael Schumacher, who also beat Massa 11-3 in 2006; only Fernando Alonso beat Massa worse than that. López's career was relatively brief but he did beat two different multiple race winners: Jérôme d'Ambrosio and Sam Bird (the winningest non-champion in the series) both 4-2 along with beating ex-F1 driver Felipe Nasr in a one-off. Robin Frijns is clearly also one of the series's best drivers as he beat Sam Bird 8-7 and only lost to last year's dominant champion António Félix da Costa 4-5 (considering da Costa beat every teammate he has ever competed against except Valtteri Bottas, this is very impressive.) Finally, Oliver Turvey has had a 2-1 record against his FE teammates, including a slight defeat of Nelson Piquet, Jr. and a 9-0 sweep of Tom Dillmann in the 2018-19 season. While I have quibbles with most of the other lists, this seems very accurate to me except that Vandoorne is probably inflated because he was Fernando Alonso's teammate.

Highest-rated drivers to never win a Formula E race (2 or more seasons)

Since there have only been six complete Formula E seasons so far, it shouldn't come as a surprise that there haven't been a large number of drivers who have had two complete seasons to begin with: there were only nine drivers who had two seasons worth of races without a win, not enough to even form a top ten list. In some cases I kind of had to fudge, such as that of Alex Lynn who only had one full-time season but made enough starts in three other seasons running part time to add up to another full-time season in terms of the number of races, which is how I determined who to include for these lists for all series.

Since Formula E has substantially more parody than F1 or even IndyCar, most of the drivers good enough to win eventually do and the drivers above a rating of .1 here are the only ones who could be argued as reasonably great. These drivers are almost entirely noted for their accomplishments elsewhere (the World Touring Car Championship for López, Formula 1 for Heidfeld and Massa, and Super Formula and sports cars for André Lotterer and Loïc Duval.) The one exception is Oliver Turvey, whose Formula E performances probably are his career highlight, but you wouldn't know it because he has not generally had fast cars yet. However, he had a winning record against past series champion Nelson Piquet, Jr., an 8-1 record against Tom Dillmann, and an undefeated record against Ma Qinghua and has stealthily had a very impressive Formula E career to date. In his junior career, he also had a better record against Jaime Alguersuari than fellow Formula E champion Sébastien Buemi did in F1, further explaining why he actually has a higher rating than Buemi.

Highest-rated drivers to never win an IndyCar championship (2 or more seasons)

On this list, Moore, Manning, and Carpentier have been thoroughly discussed so I'll gloss over those and focus on some of the others. It should come as little surprise that Justin Wilson features prominently on this list: of all the drivers who have not won championships in IndyCar in the new millennium, he is widely considered to be the best and for good reason. However, even I was not expecting Wilson to rate higher than his onetime rival Sébastien Bourdais, but I guess I'm not that surprised. For instance, both of them drove for Dale Coyne Racing and gave the perennially-underfunded team their best championship results, but it was Wilson who gave them their first win 25 years after their formation in 2009 and also their best championship finish (6th in 2015), narrowly beating Bourdais's 7th in 2018. Wilson also clearly had a better F1 career than Bourdais, and he regularly beat A.J. Allmendinger in 2005 and 2006 at RuSPORT before Dinger was fired and immediately rehired by Gerry Forsythe, whereupon he immediately tied Bourdais in wins that year, implying maybe Wilson would have won the most races and the championship had he had a Forsythe car all season. It appears Bourdais's utter Champ Car dominance was more due to his car than because he was that much better than the rest. Regardless, Bourdais vs. Wilson is a close debate and I do think Bourdais was more consistent than Wilson from season-to-season so I could understand picking him instead (their .197 and .171 ratings are not that far apart.)

By contrast, Felipe Giaffone is a big surprise. Although his first two seasons before his injury were certainly very good, he likely would have won the 2002 Indy 500 had the lapped Dario Franchitti not thrown a nasty block to allow his teammate Paul Tracy past, and he did win a race for a Mo Nunn team that never won either with Tony Kanaan or Alex Zanardi (albeit in a much shallower 2002 IRL field than the CART fields where Kanaan and Zanardi were competing), but he always seemed like a second-tier star to me regardless (he spent a long time in Indy Lights without sniffing a championship and he actually got passed by Shigeaki Hattori for the lead once.) However, I guess I was wrong and he was more promising than I realized; I did once calculate the 2002 Indy 500, where Giaffone was very impressive, as the deepest race talent-wise in IndyCar history. That isn't not surprising because I calculated the 2001 CART season to be the deepest regular-season IndyCar field in history, and all the major drivers from that season except Cristiano da Matta, Scott Dixon, Alex Zanardi, and debatably Patrick Carpentier and Alex Tagliani if you agree with my model, were in that race in addition to the IRL regulars, some of whom were also highly rated (Sam Hornish, Scott Sharp, Al Unser, Jr., and Tomas Scheckter particularly in addition to Giaffone.)

The high rating of Alex Tagliani, which almost ties the ratings of ex-teammates and 31-time winners Dario Franchitti and Paul Tracy, is shocking to me but it just makes me feel the whole triumvirate of Moore, Carpentier, and Tagliani are sort of all causing each other to be overrated. While Moore was certainly better than Carpentier and Carpentier was certainly better than Tagliani, I think all these numbers should be lower. These three drivers are clearly extremely interdependent because Moore's most frequent teammate was Carpentier and Carpentier's most frequent teammate was Tagliani. Just as Carpentier is overrated because he tied Paul Tracy in their head-to-head even though Tracy clearly outperformed him, Tagliani is overrated because he beat Will Power 8-2 when Power was a pre-rookie/rookie in 2005-2006. Tagliani tended to dominate a lot of inexperienced drivers throughout his career (3-0 vs. Guy Smith, 5-0 vs. Nelson Philippe, 6-1 vs. Marcus Marshall, 1-0 vs. Charles Zwolsman, Jr., 2-0 vs. Jaime Camara, 2-0 vs. Jay Howard, 1-0 vs. Wade Cunningham) so this partially reflects his experience advantage: when he actually had veteran teammates (Carpentier, Bryan Herta, Justin Wilson) he was nowhere near as impressive. Having said that, Tagliani only lost to one full-time teammate in his entire career (Carpentier, although he did lose to Wilson in a partial season in 2007 when Wilson's RuSPORT and Tagliani's Rocketsports merged as RSports briefly before separating late in the season) so a higher than expected rating is understandable. As with Scott Sharp, this makes me rethink some things. I had always viewed Tagliani as mediocre and Oriol Servià as the "seemingly mediocre driver who in actuality was very good" but it turns out that I probably had it totally backward. One thing that comes to mind in retrospect is that Servià never really dominated a race while Tagliani did a few times, most notably at Mid-Ohio in 2010 as an owner-driver for his FAZZT Race Team, which was fairly unusual for the era. Back then I didn't understand why the teams rated Tagliani so highly or why Dale Jarrett referred to him in a NASCAR Xfinity Series broadcast as one of the best road racers in the world. Although I still disagree with both of these judgments, I'm starting to understand a little better where they came from.

The only surprises with regard to Colton Herta and Alexander Rossi are that they aren't higher. In fact, neither would have even made the list if I hadn't made Rinus VeeKay, Ronnie Bremer, Pato O'Ward, or Rubens Barrichello ineligible due to not having two complete seasons. Herta's 2020 however had a rating of .24 so he is clearly on a radical upswing from his minor league career (and his rookie season is not included because I decided to count his Steinbrenner Harding Racing team as a single-car team even though it was clearly an Andretti Autosport satellite and some people considered it a full-on Andretti car.) I actually do believe Herta is the preseason championship favorite in IndyCar this year. I also see him becoming the first IndyCar driver in ages to attract the interest of Formula One teams considering how successful he's already been at such a young age, particularly if he wins the IndyCar title. Felix Rosenqvist may be fairly surprising here because his IndyCar career to date although good hasn't been quite as electrifying as Rossi's or Herta's by any measure. Rosenqvist has actually mildly underachieved in IndyCar so far relative to his performance elsewhere: his seasons in Formula E and especially Super Formula were better, but I do expect him to turn it around. Tomas Scheckter is mildly surprising to me but not as surprising as Tagliani. He did have blistering speed for quite a while leading the most laps in his first two Indy 500s, having the highest lead change percentage in 2005, and actually having the second-highest lead change percentage among modern drivers behind only Josef Newgarden (and since I know Newgarden fell a little in the past couple years, Scheckter might have repassed him since I last checked.) Since the model does not consider crash DNFs, obviously that is to Scheckter's benefit and he did show signs of greatness in his best races that I struggled to see from higher-rated drivers like Tagliani, Giaffone, and even Carpentier. This does help demonstrate the problem with the eye test or even just looking at raw data out of context although I still believe those three drivers and Moore are all overrated here.

Highest-rated drivers to never win an IndyCar race (2 or more seasons)

Unlike Formula One, where many very good or great drivers never win a race because they never drove a fast enough car, that really isn't the case for IndyCar nearly as much. Although there are a few drivers like Pato O'Ward (.140), Robert Wickens (.126) and surprisingly even Marcus Ericsson (.034) who would have made this list if they had two full seasons (which Ericsson did not because he missed one race in his rookie season), it seems that almost every driver who deserves a win in IndyCar and eventually stays around long enough gets one. The three big exceptions are clearly Darren Manning, who outperformed Scott Dixon in the one time period when Ganassi didn't have cars fast enough to win along with Vitor Meira and Raul Boesel, who are both frequently cited as the best IndyCar drivers never to win a race. In the past, I usually said Boesel because he spent most of his best years with Dick Simon Racing, a team that never won a race, but now I agree with the standard consensus and think Meira was better. Although Buddy Rice beat him badly in 2004, Meira turned the tables and beat Rice and rookie Danica Patrick by a large margin in 2005. Afterward, Meira gave Panther Racing their best all-time points finish after Sam Hornish left (better than either Tomas Scheckter or Dan Wheldon were able to manage) and the best season for A.J. Foyt Racing after the CART drivers crossed over (beating anything Manning, Tony Kanaan, Takuma Sato, or Mike Conway were able to do.) Boesel's very impressive 1993 seems to be mostly due to the fact that he had legendary engineer Mo Nunn helping with his cars that year, because in later seasons when he had ostensibly faster cars, he did significantly worse, especially when he replaced Jacques Villeneuve in the car he won the championship with in 1996 and finished a shocking 22nd in points. While I do think Boesel and especially Meira deserved race wins, and both of them, especially Meira, were rated above several champions (Meira was barely distinguishable from Alex Zanardi in the model), aside from them it's hard to argue anybody really deserved a win.

What most of the rest of this list seems to argue is that IndyCar teams reject Indy Lights champions too quickly. Raphael Matos, J.R. Hildebrand, and Spencer Pigot were not by any means great drivers but they were all above average and many much worse drivers than them ended up having much longer careers. It seems part of it is that the teams expect more of recent Indy Lights champions than they eventually deliver (except for Josef Newgarden and Pato O'Ward) and part of it is that drivers on the American side of the open-wheel ladder simply don't have as much money as the Europeans and South Americans tend to do so they end up being good enough to belong in the field but not good enough to be hired without finding money, so they tend to quickly disappear. If Matos, Hildebrand, and Pigot had had a long enough career to do so (and Pigot still could) they probably could have won, but I wouldn't say it's a travesty none of them won either. If Santino Ferrucci had stayed around instead of going to NASCAR and ended up at a better team than Coyne, he probably would have won an oval race eventually but he was still at times hard to take because of his contemptuous personality and the way the IndyCar media hyped him more than many greatly superior drivers, even the likes of Josef Newgarden. The fact that almost entirely forgotten drivers like Dan Clarke, Mario Haberfeld, and Gualter Salles even make this list suggests that really almost everyone who deserves a win in IndyCar eventually gets one, except for maybe one semi-major driver a decade.

Highest-rated drivers to never win an Indy 500 (2 or more starts)

This list makes considerably more less to me than the best non-champions and best non-winners lists. Ignoring the Mark Taylor anomaly, Fernando Alonso was the highest-rated driver in the entire model and while his IndyCar career has to date been unquestionably disappointing (with him being utterly dominated by his rookie teammate Pato O'Ward in last year's event after missing the 2019 race), few would argue that he wasn't the best driver to make an Indy 500 start and not win it in recent years. Josef Newgarden is the only IRL/IndyCar champion since 1999 to never win an Indy 500 so given how all his fellow champions have done, it seems inevitable that he will win one in the future (presumably the next time Penske has the fastest cars, which they definitely did not last year when he was nonetheless the highest Chevy finisher.) Granted, that is obviously not guaranteed when considering Michael Andretti's presence on the list. Some of these drivers like Darren Manning, Justin Wilson, and Sébastien Bourdais never really drove for a team that realistically could have won, although I give Bourdais credit for trying since he did replace Tony Kanaan at KV Racing the year after Kanaan won the race. However, the only drivers to win the Indy 500 for teams that were not well above average were Dan Wheldon in 2011 and Tony Kanaan in 2013, and a case can be made they were the best oval drivers of their entire epoch and better than Bourdais, who was more of a road racer. Scott Sharp definitely did have cars fast enough to win at times in the early IRL and was just a bad driver at Indy, which is one of the reasons he is significantly underrated.

The last three drivers all do seem to be overrated. Giaffone and Tagliani have already been discussed, but Alex Lloyd was not because he did not have two full-time seasons prior to his retirement in 2011 after the Wheldon crash. Lloyd had a relatively small sample size with only 11 races and all of his teammates except Graham Rahal, who he did beat in their only race as teammates, were below average and at times very bad (he was teammates with Milka Duno, Phil Giebler, Tyce Carlson, and Arie Luyendyk, Jr.) Having said that, he did post a 10-1 record, which explains his high placement on the list even if his teammates generally were not up to snuff. Bizarrely, Lloyd's only loss was to the driver generally regarded as the worst of these drivers (Duno) and that was also Duno's only win (she went 1-27 against her teammates.) That race came at Chicagoland in 2010 when Duno finished 3 laps down and Lloyd finished 38 laps down. Many people would argue that shouldn't count since Lloyd presumably went behind the wall for some time (racing-reference.info reports that he had had a spin in that race) but I decided to count all races in which both drivers were listed as running even if one of them was clearly way off the pace, because if I took time to look at statuses and lap counts for every single teammate pair, I never would have finished this project. It is interesting that even losing to Duno in a sample size of only 11 isn't enough to keep you from having an extremely good rating here as long as the rest of your record is strong enough. Lloyd is another driver like Raphael Matos, J.R. Hildebrand, and Spencer Pigot on the previous list (and maybe even Mark Taylor when considering the overall list) who makes the case that IndyCar teams dispose of Indy Lights drivers too quickly, although it is clear that Lloyd didn't want to continue in IndyCar for safety reasons after seeing the 2011 Las Vegas wreck head on and getting caught up in it (that race also prompted Paul Tracy, Davey Hamilton, and Tomas Scheckter to retire, so it clearly had a big impact in many ways.)

Highest-rated drivers to never win a Super Formula championship (2 or more seasons)

Three drivers here have higher ratings than the top-rated Super Formula champion João Paulo de Oliveira. The top-rated driver from this series as well as the top-rated driver from Japan in general is Juichi Wakisaka, who is little-known outside of Japan but was a major superstar there, winning 11 races and three championships in Super GT alongside Akira Iida and André Lotterer, but he apparently never had fast enough cars to challenge for the championship in his Super Formula appearances. He did put up an extremely good 18-5 teammate record in his six seasons with teammates and never posted a losing record in any season, although only one of his teammates was above average and that one (Katsutomo Kaneishi) only barely. Norberto Fontana had a relatively small sample size and he seems overrated because he beat Jarno Trulli 8-2 in 1995 in German Formula 3, but neither his F1 career or Super Formula career lived up to that. Andrea Caldarelli had a similar situation as his 9-3 defeat of a Jean-Éric Vergne in a junior series (Vergne's worst loss to any driver in his career) is probably the main reason for his high rating while his Super Formula career wasn't nearly that distinguished. Mika Salo had a rather bad Super Formula career but was much better in F1. Marc Goossens and Johnny Herbert were also just stopping through and their best performances were in other forms of racing.

Björn Wirdheim was especially puzzling: he won the Formula 3000 championship in 2003, then the leading F1 feeder series by nearly a 2-1 margin in points and beat his past Indy Lights champion/future IndyCar driver and announcer Townsend Bell 8-0, but he bombed out of Champ Car (where other F3000 champions and championship contenders like Alex Zanardi, Gil de Ferran, Kenny Bräck, Bruno Junqueira, Justin Wilson, and Sébastien Bourdais were usually competitive instantly) before going to Super Formula and dominating again posting an undefeated 11-0 record against his Super Formula teammates, albeit not in cars capable of winning titles. To be fair, Wirdheim's teammates were all below average but none of them were awful (not even Yuji Ide, who is nowhere near as bad as you likely think.) However, he decided to compete in Super GT and not in Super Formula for the next decade or so and did win four races, but it doesn't feel like he lived up to the potential he had.

The three remaining Japanese drivers Koudai Tsukakoshi, Takashi Kogure, and Tomoki Nojiri are exceptions to the general rule that the European drivers in the series were better than the Japanese drivers, as they were all rated over the vast majority of series champions. Tsukakoshi's main link was to Takuya Izawa, who he beat 27-15; however, he gained an incredible amount of this because Izawa was a positive-rated driver, not to mention his positive records against Sam Bird in the Formula 3 Euro Series and Vitantonio Liuzzi in Super Formula, who were both positive as well. Although Tsukakoshi only won one race because he tended to have slow cars, he has become a major force in Super GT of late. Although Kogure was beaten badly by André Lotterer, he was basically even with fellow Le Mans winner Loïc Duval, only losing to him by a margin of 10-11, which included Duval's 2009 championship season. Immediately afterward, he dominated future three-time champion Naoki Yamamoto before beating Kazuki Nakajima's brother Daisuke 14-7, and Daisuke was actually higher-rated in the model than Kazuki was. Nojiri despite only winning three races in Super Formula so far, has specialized in tying or beating a lot of drivers well-known elsewhere. He seemed to specialize in competing against international drivers as opposed to those who never leave the country, with a 5-3 record against Hideki Mutoh, 5-1 vs. Narain Karthikeyan, 3-3 vs. Stoffel Vandoorne (which carried great weight considering how high Vandoorne was rated in the model), and 2-1 vs. Pato O'Ward in 2019 the year before he took off as a legitimate IndyCar star and beat Fernando Alonso in the Indy 500. Even though these drivers did not have fast enough cars to win titles, they are all clearly rated highly for a reason.

Highest-rated drivers to never win a Super Formula race (2 or more seasons)

As for the five drivers not discussed on the previous list, Yuji Tachikawa had a .500 or better record against seven different consecutive teammates from 1998-2007 before losing to eventual four-time Super GT champion Ronnie Quintarelli 3-6 (but he actually beat him in points!) Briton James Rossiter peaked in Super Formula when he achieved an 11-2 record vs. William Buller followed by a 2018 where he had a combined .500 record against Kazuki Nakajima and João Paulo de Oliveira, which is definitely not bad. While Kamui Kobayashi was likely the greatest Japanese Formula One driver of all time with a 24-13 teammate record including earning winning records against all five of his teammates (including Nick Heidfeld, Jarno Trulli, and Sergio Pérez), his Super Formula record is by contrast a little disappointing (he lost to Narain Karthikeyan there in 2016, who was worse and lower-rated than all the drivers he beat in F1); he's clearly only here because of his exceptional F1 record. Ryo Michigami sort of had an Alex Lloyd situation where his teammates in general were very bad but he had a strong enough record (15-9) to be highly rated anyway.

In a future column, I intend to list the top-rated drivers season by season across all the major league open wheel series (I've already calculated some of this and there are a lot of surprises.) After that, I want to eventually repeat this kind of analysis first for stock car racing (NASCAR and IROC primarily, but I will also consider minor league results for the drivers that made it there), touring car racing (where I will likely include every driver that made a WTCC, Supercars, BTCC, DTM, and/or Porsche Supercup start), and possibly even rally racing and sports car racing, although those would be more complicated. For sports car racing, I realize I could do teammate driver records by taking the average speed of drivers sharing the same car and I may eventually do that. I also do eventually intend to expand this to include all the pre-2000 drivers I do not have listed here because I know that they will be a bulk of if not the lion's share of drivers to make my top 1000 greatest drivers list.

I suspect all these things will themselves take months however and this itself took about three months to calculate. I'm in the process of finishing up my book on the history of competitive typing Nerds Per Minute with a goal of completing it by March 31, and I think because of my obsessive work on this project, I'm not going to make that deadline, so that's going to be my focus for the next couple weeks, but I definitely want to redo this sort of analysis for all the other series as well. It's definitely changed my perspective on a lot of things in terms of racing analytics. I now see how bad my top 100 lists from 2015 and 2016 were, because I lacked an proper understanding or perspective of each series's place as well as failing to properly adjust for team strength and failing to realize that driver strength doesn't have as much to do with series strength as well tend to think it does (did the CART drivers who took over the IRL from 2003-2007 suddenly get worse because they switched series? Likely no.) I hope you appreciate my analysis and I look forward to sharing more with you over the coming months.

Sean Wrona is the Managing Editor of racermetrics.com, the Webmaster of race-database.com, the winner of the 2010 Ultimate Typing Championship at the SXSW Interactive Conference in Austin, and the ratings compiler and statistician for the Mensa Scrabble-by-Mail SIG. He earned a master's in applied statistics from Cornell University in 2008 and previously digitized several seasons of NBA box scores on basketball-reference.com. He is the author of the upcoming Nerds Per Minute: A History of Competitive Typing. You may contact him at sean@racermetrics.com.

---