A Teammate-Based Stock Car Driver Rating

by Sean Wrona

Since completing the initial calculations for my open wheel driver ratings last March, I decided to run the same model for stock car racing but I was decidedly more ambitious. Unlike with the open wheel ratings, where I only included drivers who made starts after 1999 or '90s drivers whose only teammate links were already included in the model, I decided to go back through all of NASCAR history for these ratings, so they took considerably longer to calculate. Simultaneously, I largely took a break from this project over the summer to finish up my book on the history of competitive typing, which was released on September 27. After I finished that, I completed my work on this project a couple of weeks before the NASCAR season ended but decided to wait to rerun the model and write up the results after the 2021 NASCAR season concluded.

As in the open wheel model, I included all drivers who made at least one start in the NASCAR Cup Series but also considered the results from any other stock car divisions where those drivers were teammates with other drivers who made Cup starts. For the most part, this included Xfinity and Camping World Truck Series races, but I also considered races in NASCAR's Convertible Series, regional touring series, ARCA, ASA, and the USAC Stock Car championship as long as I was able to determine races where two drivers drove for the same car owner and were both running at the finish of the race. For the most part, I used racing-reference.info to determine NASCAR and ARCA teammate relationships, but in many cases, I also used Ultimate Racing History to determine teammate relationships, especially for earlier years of ARCA and the USAC Stock Car Series. In those decades (especially in the '70s and '80s) most drivers in NASCAR and all other stock car divisions drove for single-car operations and multi-car teams were rarely seen in stock car racing, but I do feel that I obtained valuable information from this, especially in the 1960s, when the USAC Stock Car Series was arguably at its height and wasn't much less prestigious than NASCAR, as many of the big open wheel stars competed there, and the USAC Stock Car stars usually also made NASCAR stars as well.

Just as with the open wheel model, this model is designed to produce career average ratings for all drivers where a rating of 0 indicates a career-average driver, while above average drivers should be rated above 0 and below average drivers should be rated below 0. A rating of .100 would indicate a driver who would be expected to beat a career-average driver 60% of the time on average, while a rating of -.100 would indicate a driver expected to beat an average driver 40% of the time and so on. Once again, I only included races where both drivers finished to better evaluate drivers' pace and to remove the luck factor as much as I could. While I did include non-points exhibition races in my open wheel model because a lot of them (non-championship F1 races, the Macau Grand Prix, and so on) are actually very prestigious, I chose not to do that here since I can't think of any stock car non-points races that are held in the same esteem, and the NASCAR Clash and All-Star Race have usually been too random in terms of eligibility and officiating compared to regular races for me to feel comfortable including those in the model. I also initially planned on covering IROC in the model because I thought it might give me a lot more valuable information about drivers particularly from the '70s to the early '90s who did not have many teammates (for the most prominent example, Davey Allison had no teammate relationships so could not be included in the model, but he would have scored very, very highly if I included IROC races.) However, it seemed that many of the IROC drivers were being dragged down by their comparisons to the inexperienced open wheel, sports car, and sprint car drivers they competed against and tended to be rated lower than they should. Unfortunately, removing this element from the model didn't help much because I tend to find that almost all '80s and '90s drivers are significantly underrated by the model (unless they drove for Roush or post-'90s Hendrick, the two teams that did have consistent, long-term teammate relationships in that era) so I'm not sure it helped much.

I tried to use common sense as much as possible in determining which drivers were teammates. I know in NASCAR it was and still is extremely common to have drivers who effectively drive for the same team listed as driving for separate car owners. For example, when Roger Penske merged his team with Michael Kranefuss's in 1998, everyone knew that both those cars were de facto Penske cars and out of the same shop, and Jeremy Mayfield came out of almost complete obscurity to lead the points. Even though Mayfield technically drove for Kranefuss, I counted it as a Penske car. Similarly, I counted Kyle Petty's #44 car in the late '90s as a second Petty Enterprises car even though technically the pe2 team he owned and the Petty Enterprises team owned by his father were separate teams. I also counted the Wood Brothers car as a Penske car starting in 2015 when Ryan Blaney drove it because it is run out of the main Penske shop. I admit I had some misgivings about doing this but I was talked into it by a few people on one of the auto racing Discord chat rooms I visited. However, I noticed in the early iterations of the model that Brad Keselowski and Joey Logano were coming out too low because most of their comparisons are with each other and most of their teammate comparisons prior to that were before they had fully developed as drivers. Keselowski had been utterly dominated by more experienced teammates at the top of their game at Hendrick and Penske before Kurt Busch was fired, while Logano likewise had been at Gibbs, which meant that they each had few comparisons to drivers other than each other in their prime period, which tended to underrate both of them too much. I thought this might help correct for that, but you can easily argue here that two wrongs don't make a right. Even after doing this, Logano is still considerably lower in the model than you might expect. However, except for the Penske/Wood Brothers connection, I generally did not include satellite operations as teammates even if the drivers themselves did. I know when Stewart-Haas was a Hendrick satellite, Jimmie Johnson and Tony Stewart were always calling each other teammates in the press, but their teams were out of separate shops and added up to more than four cars in the era where teams were limited to four cars. I don't quite consider that to be the same situation, especially considering Stewart-Haas switched to Ford and was no longer a Hendrick satellite at that point, thus I counted Hendrick and Stewart-Haas as separate teams.

Below I list all the eligible drivers who had 10 or more single-race teammate comparisons. As with my previous model, I will display the results after the first iteration as well as the results after thirty iterations, when the model had largely reached equilibrium (apart from a significant underrating of most '80s and '90s drivers, which I will explain later.) Unlike with the open wheel model, I did not split the ratings by series this time as there is only one stock car series that has endured that could reasonably be considered to be a major league, while there are four such open wheel series (Formula 1, Formula E, IndyCar, and Super Formula.) I also decided splitting the calculations in this way for the first iteration was kind of a waste of time when the final iteration is more important in the first place. This time, I list drivers ranked by rating but also list each driver's career record across all their teammate relationships along with the number of shared races (defined as teammate comparisons with Cup starters in races where both drivers finished; a driver with multiple teammates does have multiple teammate comparisons if more than two drivers driving for the same team both finished.) The calculations for both the first iteration and the final iteration are complete through all 2021 NASCAR seasons.

One of the most obvious things you will notice is that the model is exceptionally kind to both '50s and '60s drivers and to a lesser degree 21st century drivers for the most part but harsh towards drivers who primarily raced in the '70s, '80s, and '90s. This is a bias that persisted throughout all iterations of the model and I do understand where it comes from. Most of the top-tier drivers of the '50s and '60s raced for premier multi-car teams at some point during their heyday (whether they were Petty Enterprises, Carl Kiekhaefer, Pete DePaolo, Hugh Babb, or Holman-Moody), which means their ratings are often very accurate. The same goes for the 21st century drivers who had even more teammate comparisons, although there are some extremely weird outliers on the earlier iterations.

However, most of the baby boomer-era drivers are way off for a variety of reasons. When these drivers came of age between the late '70s to the early '90s, there were generally extremely few multi-car teams. By the time multi-car teams again became commonplace in the mid-'90s, these drivers tended to be in their last period of relevance or beginning to decline. For example, Dale Earnhardt never had a teammate until 1997 and he was never as strong after his 1996 Talladega wreck as he was before. Ernie Irvan likewise never had a teammate until his post-injury seasons. Rusty Wallace never had a teammate until 1998 and by that point, he had already significantly declined from his peak. Darrell Waltrip especially screws things up because he got absolutely demolished by Jimmy Spencer in the 1999 and 2000 seasons and these seasons had great weight in determining his rating since there were more races then and cars were generally more reliable. Although Waltrip did have several other teammates while driving for Junior Johnson and Rick Hendrick, such as Neil Bonnett, Geoff Bodine, Benny Parsons, Tim Richmond, Ricky Rudd, and Ken Schrader, the first three of those drivers had relatively few links to drivers other than Waltrip or each other, and Schrader is also hurt because he had many bad seasons in the last decade of his career. This in turn dragged down all of these drivers on later iterations as well as their teammates. Dale Jarrett ended up negative on the final iteration largely because Rudd was one of his significant teammate links and Rudd ended up below average largely due to his comparisons with Waltrip and Jarrett. Harry Gant ended up below average because he did not beat Phil Parsons as badly as Phil's brother Benny did, and Benny was dragged down by his link with Waltrip. As a result of all this, almost all boomer drivers were significantly underrated (and some of the previous generation's drivers like Bobby Allison, David Pearson, and Cale Yarborough largely due to small sample sizes as well) so it's hard to take seriously any of the ratings of drivers who peaked in the last third of the 20th century except for Mark Martin, Jeff Gordon, Jeff Burton, and Bobby Labonte who had a ton of teammate comparisons and most of their teammate links were with drivers who came after them in eras when all the drivers had more accurate ratings (unlike for instance Dale Jarrett who was mostly linked with drivers who became stars before him in his prime years.)

While I'm significantly disappointed by the underrating of '70s, '80s, and '90s drivers I do understand it because these drivers tended to only have teammates when they were in decline while they drove for single-car teams at their peak. However, I really like the model when it comes to evaluating '50s, early '60s, and 21st century drivers relative to each other. Regardless, there are still some bizarre outliers on the original list and most of these are modern drivers. Some of these like Ross Chastain and John Hunter Nemechek are not that shocking because everyone knows that in the minor leagues especially they've been dominating all their teammates for years (Nemechek even had a winning record against Kyle Busch in the trucks this year even though he didn't win the title), but a few of them are genuinely shocking and foremost among these are Ryan Sieg, Cole Whitt, Corey LaJoie, and Chris Buescher.

Sieg is a driver who has always been incredibly difficult to evaluate because like his newly-famed contemporary Brandon Brown, he has spent almost his entire career driving for an underfunded family operation, but he did beat his teammates by more than a 2-to-1 ratio, which not many others did and this included winning records against Chastain and Jeff Green and an 11-0 sweep of David Starr. LaJoie also had a greater than 2-to-1 ratio against his teammates, including being the only driver to sweep Sieg 2-0 along with beating all his semi-regular teammates Gray Gaulding, Brett Moffitt, and Justin Haley by 2-to-1 ratios or more. Buescher benefited from his Xfinity championship season, where he thrashed Elliott Sadler and Bubba Wallace harder than a lot of other drivers did, as well as his recent seasons where he got lucky to compete against Ryan Preece as a rookie and Ryan Newman in his post-injury decline period. The big flaw of this model is that it assumes that drivers' abilities remain consistent throughout their career when obviously everyone knows that isn't true. Hence drivers who blow out good teammates in their bad teammates will be arguably unjustly rewarded while drivers who struggle against bad teammates in their better seasons will be penalized, but I do think that's how it should be and overall the idea of the model is more good than bad. Having said that, maybe it would improve if I somehow threw out rookie seasons or post-injury seasons (or alternatively, split drivers whose careers significantly declined after injuries like Newman, Sterling Marlin, Steve Park, or Tony Stewart into two separate careers before and after their career-altering injuries.) To be fair to Buescher, he did also have a 3-0 sweep of Chastain in 2020 as well as a 2-0 sweep of Carl Edwards as early as 2012, while also being the only driver to have more than five races against Cole Whitt to tie him, so I do think Buescher is probably legitimately underrated but probably not by as much as my model will tell you.

But it is Whitt who I find most inspiring. I had already spotlighted him in 2019 when I did my first NASCAR teammate analysis and noticed that he trounced every teammate that he competed against in five or more races by at least a 4-1 margin except for Chris Buescher and (weirdly) David Gilliland, who perhaps as a result of this ends up way higher in all iterations of the model than you might expect (still negative, but not by a lot, when he is often regarded as one of the worst long-running drivers in Cup history.) Most notably, Whitt defeated Alex Bowman 18-2 when they were both rookie teammates in 2014, which actually resulted in him being the highest-rated driver in the model that year, just as Buescher was for 2021 when you consider how badly he beat Newman relative to Newman's usual performance. While I don't think Whitt was the best driver then and I don't think Buescher is the best driver now, I at least do get where both of those statistics come from. Having said that, it is clear that while he may still be the worst Hendrick driver, Bowman isn't a hack no matter what Denny Hamlin thinks and he has significantly improved since 2014, so Whitt is probably being overrated to some degree because Bowman wasn't as good then as he is now, much like Buescher is probably being overrated because Newman has fallen off considerably from his prime years. Having said that, I do think it's clear that both of these drivers (especially Whitt) deserved more than they got, although Buescher certainly has an opportunity to test his mettle now that he will be teammates with a past champion in Brad Keselowski for 2022.

Drivers like these were generally overrated in the original model because they tended to absurdly dominate really bad drivers who themselves significantly declined on later iterations of the model. As with the open wheel model, drivers who had strong teammates continued to increase on future iterations while drivers who had weak teammates declined. This didn't work for the '70s to '90s drivers, however, because a lot of drivers were incorrectly marked as weak by the model due to having only their weaker late-period seasons counted or their connections to drivers like Darrell Waltrip who were incorrectly marked as bad. The model views Waltrip more as "the guy who got thrashed by Jimmy Spencer" than anything else because he didn't have teammates in many of his best seasons, and so a very good driver like Neil Bonnett because he had few connections outside of Waltrip is viewed as "the guy who got thrashed by the guy who got thrashed by Jimmy Spencer", and ends up being extremely underrated like all the drivers of his era when all was said and done. However, for the 21st century drivers, I tended to find the later iterations very accurate in evaluating their overall talent. The oddities of Sieg, LaJoie, and Buescher do correctly drop considerably when accounting for the fact that their bad teammates were overrated in the initial iterations of the model, but they do still remain above average drivers even after iterations were complete, which means they are still likely significantly underrated by fans.

Whitt, however, barely dropped at all from the first iteration of the model to the thirtieth iteration. While many of Whitt's bad teammates declined and dragged him down, Bowman rose significantly thanks to the strength of his teammates to compensate for this. Even after iterations were complete, Whitt is still rubbing shoulders with Cup champions and megastars - he even ended up with a higher overall career rating than Tony Stewart! Perhaps this shouldn't be shocking because Whitt was something of a child prodigy, becoming the youngest-ever USAC Midget champion when he was still a teenager, and extremely early success is often correlated with greatness. Having said that, Josh Wise was one of the youngest ever USAC Sprint Car champions and he was very low rated in all iterations of the model, so early prodigy alone guarantees nothing. Having said that, I'm now starting to think Whitt may have actually been the best driver to never win a Cup race. Sadly, he's now best known as "the guy with the most races without a top ten finish" even though it's clear that drivers in the kind of cars he had could not get top ten finishes unless they got extremely lucky in plate races, and it just never happened for him. I think he just might be the most underrated Cup driver ever.

Aside from those drivers, the top ranks of the first iteration of the model are largely made up by drivers who are already established legends, even if some of them don't get the respect they deserve. Lee Petty had an overwhelming advantage in leading the list with a career rating of 0.288, indicating that he would beat the average driver 78.8% of the time. He significantly benefited from being teammates with his son Richard during his first couple seasons before his injury at the 1961 Daytona twin 125 qualifying race effectively ended his career. Lee badly thrashed Richard 31-11 in those years because Richard had not yet fully developed. It seems inevitable that Richard would have overtaken Lee and reversed this at some point, but the injury prevented that from happening so Lee ended up claiming first place by a large margin. Most of the rest of the drivers in the top forty will not come as much of a surprise, but a few will. Paul Goldsmith did win 9 NASCAR races in a brief career, but never got the respect he was due because most of his best results came in the rival USAC Stock Car series that has seemingly been erased by history, even though it wasn't that much less prestigious than NASCAR at the time. Goldsmith won back-to-back championships in that division in 1961 and 1962, setting a single-season win record there that would never be broken by winning ten races there in 1961. Goldsmith did have plenty of big names as teammates in both USAC and NASCAR, and had at least a .500 record against all 17 of his teammates except for Bob Welborn; this included having winning records against big names like A.J. Foyt, Bobby Isaac, Junior Johnson, Fireball Roberts, and Don White, the all-time USAC Stock Car win leader. Goldsmith was clearly without question one of the best stock car drivers of the '60s but doesn't get his due now because most of his best results weren't in NASCAR.

The same goes for forgotten legend Roger McCluskey. McCluskey was the only driver other than A.J. Foyt to win USAC Indy Car, Stock Car, and Sprint Car titles in his career but tends to get short shrift nowadays. One of the reasons is that his 1973 IndyCar championship season is widely regarded as one of the worst ever: he only won one race at Pocono due to attrition and largely benefited from avoiding DNFs in an era where they were rampant and the fact that the superspeedway races carried more weight since points were determined in part due to race distances; it didn't help that he only won five races in his career, one of which came in 1979 in his final start against an extremely shallow USAC field when all the top drivers other than A.J. Foyt were racing in CART. Considering this and the fact that the USAC Stock Car Series has seemingly been erased from history despite its prestige, McCluskey doesn't get the respect he should today despite his great versatility. His stock car results are arguably his most impressive, however, as he won 23 races in that division along with the 1969 and 1970 championships, where he utterly dominated his only teammate Norm Nelson 20-3, and Nelson was a great driver in his own right as he won a Cup race in only five starts and won three USAC Stock Car titles himself, although I think by that point Nelson had significantly declined due to age. Regardless, McCluskey does well throughout the entire model and in its final iteration, he takes a commanding lead over Lee Petty to become the highest-rated driver in the model.

Some other drivers here are worth noting. People tend to remember Tim Flock as an all-time legend but ignore that his brothers were held in a similarly high regard. Fonty Flock had the best season of 1951 by most account sand outperformed his brother that year. Tim was certainly better as he doubled Fonty in wins as well as beating him 8-4 in shared teammate races, but both of them were remarkable as Tim posted at least a .500 record against every single teammate and Fonty went undefeated against every teammate except Tim, including the likes of Buck Baker, Speedy Thompson, and Curtis Turner. Although neither of them have extremely high sample sizes, both Tim and Fonty remained in the top five constantly throughout the model. Analytics generally are exceptionally good to Fonty and prove that it was clearly a mistake that he was excluded from NASCAR's original 50 Greatest Drivers list (cut out either Glen Wood or Ralph Earnhardt.) Although Fred Lorenzen was always regarded as a legend, I didn't notice how much more dominant he was than his other Holman-Moody teammates until I did further research; his records of 7-1 vs. Dick Hutcherson and 10-1 vs. Nelson Stacy are very impressive and his 8-5 record vs. Fireball Roberts was perhaps even more so. I used to think Roberts or Junior Johnson was the best non-champion of that era, but I think I was wrong. It was clearly Lorenzen or Goldsmith. Admittedly, both of them were multi-time champions in the USAC Stock Car Series.

First Iteration

Final Iteration

Just as in the case with the open wheel model, the drivers who had the strongest teammate averages (who naturally tended to be the strongest drivers as the top teams attract top drivers) rose while the drivers who had the weakest teammate averages (who tended to be weaker) fell. As I already mentioned, the '70s-'90s drivers tend to take a hard hit since most of these drivers did not have teammates in their prime years but did in their decline years. For those drivers, I tend to think the ratings on the initial list are closer to accurate, but I definitely like this ranking more for the drivers of the '50s and '60s, as well as drivers who had most of their relevant period in 1995 or later.

The model really liked '50s and '60s drivers in all iterations, as the top 21 drivers to gain the most between the 1st and 30th iterations all came from this time period; the driver not from this era who gained the most in later iterations was Chase Elliott in 22nd, who improved from .126 to .220, but most of his Hendrick teammates were not far behind. The ten drivers who improved the most were Roger McCluskey (.318), who went from .203 to a mind-boggling .521, Jim Hurtubise (.278), who went from a deeply negative -.157 to well above average at .121, Curtis Turner (.189), Tim Flock (.172), Fonty Flock (.163), Bob Welborn (.155), Lee Petty (.147), Bobby Isaac (.142), Norm Nelson (.125), who basically benefited from McCluskey's rise despite getting destroyed by him, and Jim Paschal (.118). Most of these drivers were linked in teammate relationships with at least one of the others, so they tended to all propel each other upwards faster than the modern drivers did. I think this is likely because modern drivers already had a large number of teammate relationships, which tended to moderate them while drivers with small sample sizes are better able to make wild swings. Additionally, I would speculate that in that era, most of the drivers on multi-car teams were good while few bad drivers made starts on multi-car teams at all, meaning for the '50s and '60s drivers, bad drivers are largely removed from the model entirely, which tended to inflate all the drivers from this era more.

The drivers who fell the most came from a variety of eras. These were Earl Brooks (-.312), who dropped from -.148 to -.460, followed by Phil Parsons (-.311), Wendell Scott (-.279), Shep Langdon (-.251), Chad Finchum (-.245), Terry Cook (-.233), George Green (-.231), Matt Mills (-.224), Don Gray (-.223), and Eric McClure (-.218). The common thread here is that most of these drivers either came from the same '50s/'60s era which allowed wilder fluctuations or primarily raced in minor league series in the 21st century against well-below average teammates who also fell significantly throughout later iterations of the model, as was the case for Finchum, Cook, Mills, and McClure. Two of these drivers (Finchum and George Green) actually had positive ratings after the initial model at 0.020 and 0.006 respectively, but fell to -.245 and -.231, which does seem closer to the truth. Phil Parsons was collateral damage from the Waltrip factor, as Benny Parsons was an important link for both of them so they all steadily fell together (I do think Phil was below average, but nowhere near that bad.) I do feel bad that Wendell Scott fell that much, but it's understandable because he was 5-2 against Earl Brooks, who declined more than any other driver in the model, and got swept 4-0 by Jabe Thomas, who also fell significantly from 0.025 to -0.164.

The model certainly does make more sense after iterations were complete. All 20 drivers rated above .2 are certainly inarguable legends, as are most of the drivers who fell in the range of .1 to .2 as well. While you might want to argue Paul Goldsmith and Fonty Flock are not legends, I'm afraid I must disagree. Ron Fellows, who was another surprise on the original model with a rating over .2, did actually continually improve and in fact trails only Kevin Harvick and Jimmie Johnson among 21st century drivers. However, that's not necessarily spurious as Fellows did have a 19-7 record in his various road course starts across all NASCAR divisions, including 2-0 sweeps of Dale Earnhardt, Jr., Brad Keselowski, and Cole Whitt (and recall how much my model likes Whitt, who actually finishes with a higher rating than both Junior and Brad.) Fellows may never have won a Cup race, but he was clearly an absolute top-tier road course talent for a very long time in his Cup career in a way that Scott Pruett and especially Boris Said weren't (Pruett was below average and Said was way below average.) I do think the wild ascent of '50s and '60s drivers generally is likely overrating some of the solid drivers who were non-champions of this period, probably Goldsmith and Sam McQuagg most, but also Jim Paschal, Speedy Thompson, Jack Smith, and Gwyn Staley, who all seem like they should be lower, but I'll stick it up for Paschal most of these, because he did only lose to Richard Petty 15-21 (and actually had an advantage on him prior to 1963) in addition to also having winning records against David Pearson and Lee Petty (in admittedly very small sample sizes.)

While I have some issues with the placement of the 20th century drivers in all eras, I think the 21st century drivers' rankings are almost completely dead on, give or take the aforementioned Cole Whitt. While I'm not sure I would take Kevin Harvick over Jimmie Johnson as the best driver of this century, I do understand where it comes from because Johnson declined in his late career while Harvick (so far) has not. Some may argue Harvick declined in 2021 because he went winless for the first time since 2009 and especially because his greatly inferior teammate Aric Almirola did win, but as far as his teammate comparisons went, he remained just as dominant as ever; it's just that his cars tended to not be fast enough for him to lead at most tracks. Admittedly, the difference between Harvick and Johnson is very slight, and if Harvick has a significant decline period like Johnson's 2018-2020 he will almost certainly drop below him. It should come as no surprise that drivers at their peak in recent years like Chase Elliott and Kyle Larson will feature prominently since they have not yet declined, but I'm a bit surprised after Larson's domination of Elliott this year that Elliott is still ahead. However, one year does not make a career, and even though Larson beat Elliott significantly this year (posting a .303 rating for 2021 compared to Elliott's .138), Elliott did beat him all three years from 2017-19 to compensate, including being the highest-rated driver in the model in both '17 and '18, although those were likely inflated because Elliott was competing against a washed-up Dale Earnhardt, Jr. and Kasey Kahne in 2017 and against the inexperienced William Byron and Alex Bowman the following year.

The major Roush drivers all ended up extremely close in the ratings, with Mark Martin and Matt Kenseth both posting career averages of .185 (although Martin was slightly ahead by less than 2 ten-thousandths), Carl Edwards at .183, and Kurt Busch at .172. All these drivers rated behind Kyle Busch (.202) but ahead of Denny Hamlin (.165) and that seems largely correct to me. Kenseth narrowly had winning records against Edwards (199-188) and Hamlin (72-65), but Hamlin did beat Edwards (32-26), so he may be a little underrated here. However, Hamlin was ahead of Martin Truex, Jr. (.151) which is also not surprising. Hamlin was much more consistent from season-to-season than Truex was, as Truex had several inexplicably bad seasons in his early career. Additionally, Hamlin has a slight advantage over Truex (56-44) in the years they have been teammates.

Among the modern drivers, Dale Earnhardt, Jr. led the way among the second tier of drivers with a career rating of .126, but he did beat his buddy Tony Stewart who had a surprisingly low rating of .120. Both Junior and Stewart are probably underrated in this model because their careers were riddled by injuries. The turning point in Stewart's career seems to have been the sprint car crash in August 2013 where he broke his leg; it was likely the moment Stewart switched from being good to bad, although admittedly he wasn't very impressive in 2013 prior to that as he lost to Ryan Newman 4-12 that year even before the injury. Nonetheless, if I only calculate Stewart's teammate comparisons through 2013, his rating rises to .180, which is a lot more realistic. Since each teammate is judged by their teammates' career average ratings, if I used that for Stewart's rating instead, all his teammates would also rise, which would likely raise Stewart's rating beyond that. However, I suspect that his rating would only be around Kyle Busch's rating at the absolute highest, and he would probably still be behind Jimmie Johnson, Jeff Gordon, and Kevin Harvick either way, even though he was regarded by many as the best driver of his time. The other major drivers in this tier included such names as Buddy Baker (.123), Jeff Burton (.120, which is much higher than I expected), Brad Keselowski (.118), Ryan Newman (.111, also higher than I expected), Greg Biffle (.111), and Dale Earnhardt (.107), who fell considerably as his teammate Mike Skinner did. However, when you consider Earnhardt is being solely decided by his decline years of 1997-2000, that is still a high rating, since it seems like having a rating of >= .1 seems to be roughly equivalent with greatness. This means if he'd had a teammate in the '80s and early '90s, he'd have to have a rating of .2 at the very least.

Even after counting the recent Wood Brothers entries as Penske cars, neither Brad Keselowski nor Joey Logano rose as high as you might expect despite being champions. I did not expect Keselowski and Logano to be roughly equivalent/slightly lower rated than Jeff Burton and Clint Bowyer, but I suspect Burton and Bowyer were partially lifted up just because of how high-rated Kevin Harvick is, while as earlier mentioned, Keselowski and Logano struggled when they had teammates other than each other. The ratings for Keselowski and Logano do not seem to be spurious because they did well against each other but generally struggled outside of their Penske years. They seem to legitimately be overrated, although I still think my model is underrating all the Penske drivers as Ryan Blaney is barely above average and Austin Cindric is far below average, which both seem wrong.

Most of the drivers who had unexpectedly high ratings after the first iteration did decline significantly as they were clearly initially overrated by dominating weak teammates who continued to decline on further iterations, but none of them fell below average. As earlier mentioned, Cole Whitt barely dropped at all (.152 to .142), but most of the others dropped considerably: Ross Chastain dropped from .166 to .078, Chris Buescher from .107 to .064, Kevin Swindell from .125 to .044, Corey LaJoie from .136 to .042, Ryan Sieg from .193 to .039, and Brennan Poole (inflated by dominating his fellow Ware drivers in 2020) dropped from .097 to .016. These are/were clearly all good drivers, but few of them could yet be argued as great. I suppose Chastain and Buescher may still have potential.

One big surprise was Robby Gordon, who actually slightly improved from .049 to .052, slightly ahead of Bobby Labonte but slightly behind Ward Burton, to cite two of his contemporaries. He too I suspect mostly rose because of how high-rated his onetime teammate Kevin Harvick was (and he still remains the only Harvick teammate to win more races than him in his career), but the biggest surprise to me by far was that Robby was the highest-rated driver of 1997, even as a rookie. Surprisingly, this result does not seem to be spurious as he beat Joe Nemechek 8-2 and Wally Dallenbach 3-0, basically roughly equivalent to what Jeff Gordon was doing when he was their teammate. Granted, Robby crashed a lot as a rookie and I threw DNFs out in the model, but it seems that Robby did have some potential for greatness in his rookie season, but his switching from one series to another blunted his momentum and he never lived up to the potential of that first season. Regardless, his generally acknowledged talent and car control did shine through to some degree on this list.

While I do think almost all drivers from the mid-'60s to 2000 are underrated by this model, and the drivers of the '50s and early '60s are probably overrated, I do really like how the model came out for the 21st century drivers and I think it came out very well. Perhaps the model would be even more accurate if I simply threw out all drivers' teammate comparisons prior to 1992, the start of Jeff Gordon's career and Mark Martin's career at Roush. Perhaps I will do that sometime. The main flaw may be that this is in its essence a consistency-based model rather than a dominance-based model. If you don't believe luck necessarily evens out in the end (which I don't, and even less now when the modern championship format is more luck-dependent than any of the previous formats, not that I think those championship formats were very good either) perhaps a model where I determined teammate wins based on which teammate led more laps would be more accurate in ranking drivers, particularly when it comes to earlier eras when there was a lot more attrition. I suspect the results of both a model like this and a model like that would be very similar though.

The other thing I may have to reconsider in further versions of this model is whether to split drivers' careers in two based on injuries or other career-altering events. I have already stated that Tony Stewart's rating would be at least .180 (probably higher) instead of .120 if I threw out his post-injury seasons of 2014-16. The starkest such difference is with Steve Park, whose rating through 2001 is .193 while his career rating was -.139. Clearly, very few drivers were as affected by injury as Park was, and I suspect this is the main reason why most of Dale Earnhardt, Jr.'s DEI seasons turn out to have lower ratings than it seem like they should. Similarly, Sterling Marlin's rating would be at least .033 if I stopped at 2002 instead of -.024 if I include his later seasons. I even think .033 is still underrating Marlin but I suspect he is also badly affected by the underrating of '90s drivers (Bill Elliott and Geoff Bodine are far worse here than they should be, probably both side effects of the Waltrip/Spencer issue) Clearly, these three drivers especially were significantly different before and after their injuries. The same can be said for Waltrip himself, as the real turning point in his career seems to have been when he was injured in the 1995 Winston All-Star Race; although he had declined, he was still pretty solid in 1993, 1994, and the early portion of 1995, but he was pretty much washed up afterward. If I split all these drivers' careers in two, I suspect a lot of the ratings for '90s and early 2000s accurate might be more accurate.

Since the model is designed so that the average driver is 0 and so that most drivers are rated between -.5 and .5 just as in the case of the open wheel model, the ranges of ratings in this model will generally correspond to the ranges of ratings in the open wheel model as well. Hence, in general, drivers with a rating >= .2 can be considered career-long elite, drivers >= .1 and < .2 great, drivers >= .05 and < .1 very good but not great, drivers >= 0 and < .05 good, drivers >= -.05 and < 0 are on the good side of mediocre, drivers >= -.1 and < -.05 are on the bad side of mediocre, drivers >= -.2 and < -.1 are bad but not awful, and drivers < -.2 are very bad. But you can certainly still take many of these ratings with a grain of salt, especially drivers who have fewer than 100 teammate comparisons prior to the 21st century. For the current century, I do think they are very accurate.

In the open wheel model, I listed the top drivers who never won individual races or championships in each of the four major league open wheel series. This time, I won't be doing that. Instead, I'm going to list the top five highest rated drivers from each year and discuss some general trends among the year-by-year data. I have calculated the complete ratings for all drivers in the model who had ten or more shared teammate comparisons for every year from 1990 to 2021, and I will likely print them in future columns for the rest of the year, but I will only be listing the highest-rated drivers for each year here.

One of the oddities of the model is the fact that a driver can actually have a winning record against a teammate but appear lower in the seasonal rankings. In 1991, Sterling Marlin solidly outperformed his Junior Johnson teammate Geoff Bodine, finishing 7th in points to Bodine's 14th and beating him 10-8 in shared finishes. However, because Bodine's overall rating is significantly lower than Marlin's (since Bodine is skewed by the Waltrip effect) Marlin was expected to beat Bodine by more than he actually did, which surprisingly resulted in Bodine being rated higher. The same thing happened in 1996 when Jeff Gordon is rated below Terry Labonte despite beating Labonte in every single category except consistency. Despite losing the title, Gordon did have a winning 13-11 teammate record against Labonte, but considering how many of Labonte's bad later seasons were included in the model, Labonte's overall rating drastically understates his ability and Gordon was expected to beat him by more. Weird things like this did happen periodically so that is definitely one of the model's main liabilities.

There are many problems with assuming equal performance throughout a career as this model does. For one, drivers who had significant injuries (especially late in their career) tended to have wildly different levels of performance before and after the injury, so a rating over the entire career will underrate the drivers' teammates' performance prior to the injury and overrate it afterward. Even though most people would argue Bill Elliott outperformed Mark Martin in 1992 (I would too, but I do have to think about it slightly since Martin did beat Elliott in lead shares), he is rated below Martin because Sterling Marlin's average rating is not a proper measure of his ability at that point. If I stop at 2002 and use Marlin's pre-injury rating of .033, Elliott's rating rises to .200, which more closely matches the conventional wisdom. Similarly, if I use Stewart's pre-injury rating of .180 instead of his career rating of .120 for his rookie season of 1999, Bobby Labonte's rating would rise .060 to .260, which seems closer to correct to me since I think Labonte was the best driver that year.

As I've already mentioned, the model also tends to overrate drivers who beat good drivers when they had unusually bad seasons. The most obvious examples of this are Kurt Busch setting the highest season rating of 0.474 last year thanks to beating a washed up Matt Kenseth 21-4, Chris Buescher leading this year by beating a washed up Ryan Newman 21-8, and Cole Whitt actually leading the overall ratings for 2014 due to his utter domination of fellow rookie Alex Bowman. Many of the most overrated seasons seem to happen when drivers are teammates against rookies or generally inexperienced drivers. For instance, Dale Earnhardt, Jr. leads for 2006 because he dominated his rookie teammate, eventual champion Martin Truex, Jr. and he did fairly well in 2007 also because he did actually win the head-to-head against Truex despite only Truex winning a race and making the chase that year. The DEI drivers are generally overrated in this period, which also includes Mark Martin and Truex in 2008 because Paul Menard was in his second season and Regan Smith was a rookie, which easily allowed them to dominate and inflated the strength of their seasons. Jimmie Johnson and Jeff Gordon both set among the highest overall scores in 2004 because they were teammates to a washed-up Terry Labonte and a rookie Brian Vickers, but you can still certainly make the case they were the top two drivers that year although I would probably take Dale Earnhardt, Jr. (curiously underrated in this period) over Gordon. But despite performing weaker, Junior's late period is strangely overrated because he benefited from a washed up Jeff Gordon and Kasey Kahne and a rookie Chase Elliott, even taking a commanding lead in his part-time 2016. Chase Elliott's 2017 and 2018 may also be inflated because in the former year he was benefiting from a washed up Junior and Kahne and an inexperienced Byron and Bowman, respectively, although I do think his 2018 was basically as strong as anything else when you adjust for his equipment and how weak the Chevies were that year.

Most of the seasons where drivers generally considered to be mediocre landed in the top five are of that nature. For instance, Ricky Stenhouse benefited from beating a rapidly declining Matt Kenseth in his head-to-head teammate record in 2018. However, in several cases, certain teammates did simply benefit from unexpected slumps, such as Juan Pablo Montoya placing 3rd for 2009 when he dominated an unusually weak Truex and Jamie McMurray placing 2nd when he posted a winning record against teammate Kyle Larson for the only time in his career. Although everyone acknowledged Johnny Benson's 2001 was good, I don't think many people would expect it to be in the top five, but he utterly dominated Ken Schrader 25-3, which is roughly equivalent to what Jeff Gordon was doing against him in the mid-'90s. Personally, I think Schrader was probably frightened after witnessing Earnhardt's fatal crash personally and being the first person to see his corpse, so he was probably extremely careful to not take any chances that year, which is understandable. Ryan Newman's three top five seasons seem largely random. While 2003 should surprise no one, the other two years probably will. In 2013, Newman trounced Tony Stewart 12-4 even before Stewart's career-altering injury and also destroyed a washed up Mark Martin 8-3. However, he actually didn't get much of a boost from competing against rookie Danica Patrick. Her rating was so low that he actually fell in the ratings when considering his results against her, so most of that rating is coming from his comparisons with Stewart and Martin. In 2019, Newman beat Stenhouse 21-9, which is pretty much the same level of domination Carl Edwards was doing against him, so actually neither of these results is really spurious. Newman kept up his consistency a lot longer than people realized, which is why he still appears higher in the ratings than you might expect, enough for Buescher to lead the ratings this year for dominating him so badly. The year 2000 is utterly dominated by drivers who were inflated by rookie teammates, as Jeff Burton and Mark Martin got to capitalize on a rookie Matt Kenseth and part-time pre-rookie Kurt Busch, Ward Burton dominated rookie Dave Blaney, and Greg Biffle dominated rookie Kurt Busch in the truck series; only Jeff Gordon of that year's top five had no rookie teammates. To be fair, both Burton bros kept it up the next year with Burton being the top-rated Cup driver in 2001 and Ward ranking 6th, so they were generally pretty great in this period regardless. Rookie teammates or not, I still think Burton was the best driver in 2000.

My decision to include all stock car series in this model (except for IROC) is probably a controversial one, as I suspect most people doing a model like this would choose to include Cup races only. As a result, some minor league drivers do make the top five occasionally. Rick Mast and Elton Sawyer did so for their 1990 Busch season, as did Jeff Green in 1997 for his domination of the Sadler brothers at Diamond Ridge Motorsports before he was called up to Cup, Joe Ruttman for his defeat of rookie Roush teammate Greg Biffle in 1998, Biffle himself for his truck and Busch successes from 1999-2001, and Martin Truex, Jr. for his 2004 Busch season where he tied Dale Earnhardt, Jr. in their Busch starts along with utterly dominating part-time rookie teammate Paul Menard. These were all clearly impressive minor league seasons worthy of consideration although no minor league season has made the top five since. Biffle and Truex especially seemed to have been ready for Cup long before they actually started there. If I excluded the minor leaguers, Greg Sacks would be listed 4th and last for 1990, Jeff Burton would be 5th for 1997 and 1998, Dale Earnhardt would be 5th for 1999, Ward Burton would be 5th for 2001, Matt Kenseth would be 5th for 2004, and Steve Park would jump from 8th to 5th for 2000 (6th and 7th for that year are fellow minor leaguers Jeff Green and Kevin Harvick.)

The reason I decided I wanted to include minor league races was because it simply provided more data for my data set, especially for some more obscure drivers who were Cup part-timers but never had a full-time Cup teammate. However, this unfortunately means that some drivers who had substantially different performances in Cup and minor league races may have their rankings altered if they were substantially better or worse for instance in Cup than in Xfinity or vice versa. The most notable victim in this regard is Carl Edwards, who would have posted a series-leading rating of .355 for 2008 if I counted only his Cup races, but dropped to 3rd because I included his Nationwide races, where he was a lot less successful. For example, in Cup Edwards was 23-10 vs. Greg Biffle, 24-8 vs. Matt Kenseth, and 29-4 vs. David Ragan, but in his Nationwide starts, he lost to Biffle, tied Kenseth, and beat Ragan only 17-14. Clearly he was substantially better in his Cup starts and most people would say that matters more. Similarly, Kyle Busch lost to Matt Kenseth in his Cup head-to-heads in both 2013 and 2014 but beat him by a combined 26-6 in his Nationwide starts in those years, which clearly overrated Busch and underrated Kenseth compared to how it would be if I counted only Cup starts (this probably explains why Kenseth does not make the top five for 2013, but he was sixth.) While Edwards and Kenseth are the most notable victims of my decision to include minor-league starts, they aren't the only ones. In 2016, Kevin Harvick placed only 6th because he had 3-3 records against two below-average rated drivers Elliott Sadler and Justin Allgaier in Xfinity, but if I counted the Cup races only, he would rise to 2nd and 1st among Cup full-timers. Although Chase Elliott and Harvick were very evenly matched in the late 2010s, the individual seasons where they led the ratings were largely contingent on their Xfinity results. In 2018, Harvick had a 1-2 record against Cole Custer in Xfinity while in 2019, Chase Elliott had 0-1 records against Michael Annett and Justin Allgaier. If those races were not counted and only the Cup starts counted, Harvick would take the overall lead for 2018 and Elliott would take the overall lead for 2019. Clearly, counting all series is messing up the evaluation of certain Cup seasons, but I still think the additional data this decision provides adds to the model rather than detracting from it.

In the early 1990s, there were still very few multi-car teams in stock car racing. In 1990, Hendrick was the only multi-car team in Cup, although they were joined by Junior Johnson the next year and Roush the year after that. You can see how underrated this era is when you look at the driver ratings for all the drivers of this era. Obviously the fact that Waltrip was underrated by his Jimmy Spencer years led to an underrating of all the Hendrick drivers in general, so these results should be taken with a grain of salt, although I think the rankings within drivers on the same team are still usually correct. Surprisingly, despite the lack of multi-car teams in Cup that year, Rick Mast and Elton Sawyer were teammates in Busch for the Alan Dillard team so were therefore eligible for the list. Despite how low his overall rating was, Ken Schrader was a very relevant driver in this period as he led the list for both 1990 and 1991. Despite Schrader's teammate Ricky Rudd finishing second in points that latter year, Schrader was much faster on a week-by-week basis and actually beat Rudd 14 races to 9. If the data here were accurate and Rudd's rating was what it should be, Schrader's rating would probably be somewhere on the order of .2, but because of the vast underrating of this era, he and all the drivers are probably lower-rated than they should be.

Starting with the year 1992, Mark Martin had an extended run of dominance according to the model where he finished in first place six out of eight years and placed a close second the other two years. However, it's worth noting that from 1992-1994 Bill Elliott was really the only marquee name who drove for a multi-car team because the three other drivers who dominated this era along with Martin (Dale Earnhardt, Rusty Wallace, and Ernie Irvan) never had teammates in this period. Furthermore, as I earlier mentioned, if you only consider the pre-injury portion of Sterling Marlin's career, Elliott would certainly overtake Martin for that year. However, it is the late '90s that interests me more. Despite dominating this entire period, Jeff Gordon only ranked first in the model one time in 1995 (barely over Martin) and never did so again, although he did rank second in seven future years. Even in 1998, Martin narrowly edged out Gordon, although it was certainly very close. It's worth noting that Martin regularly bested Gordon and everyone else in the late '90s while competing in IROC and 1997's top-rated driver Robby Gordon finished second to him in the IROC standings. Although I initially intended to include IROC in the model, I eventually decided not to so none of the results from that series are included in this analysis. Most people considered IROC to be a novelty exhibition of no real importance, but this does raise the question - was Martin actually consistently better than Gordon in this period? I still say probably not because it seemed like the Fords were consistently faster than the Chevies from 1997-2000: for instance Chad Little fought for a race win for Roush in 1998 and none of the drivers in the Hendrick #50 car did. I suspect the reason Gordon is usually rated under Martin is because Terry Labonte is generally underrated by the model because he had many bad seasons to finish his career, but so did Martin's primary teammate Jeff Burton to a lesser degree. Regardless of what my model says, I do probably think Gordon was still better, but I think Martin was probably the best driver at some point in the '90s.

Martin keeping it up as late as 1999 is surprising as it seemed by that point that Jack Roush was stealing most of the headlines for Roush, but Martin did win their 1999 head-to-head narrowly (15-14) in Cup and by a larger margin (6-2) in Busch. However, after that Burton clearly took over control of the team the next two years and led Cup both seasons in the yearly ratings. Burton's 2001, like Sterling Marlin's 1998, Ward Burton's 2000, Kevin Harvick's 2004-05, and Martin Truex, Jr.'s 2011 seems to genuinely reflect something close to truly elite performance in a subpar car. The season it is probably most similar to is Harvick's 2021 in that Burton continued to dominate his teammates as usual but simply had slower cars to do it with, which made his year look significantly worse than it was. I like that the model does seem to correctly identify seasons like this where drivers achieved great performances in second-rate cars, which is definitely a strength of the model. Although lots of drivers make random appearances here and there, the same drivers tend to make the top five repeatedly from year-to-year, and often even in the same positions, another strength.

The 2000s were generally a remarkably even period and generally the deepest era in Cup history. How deep? In the ten years from 2000 to 2009, eight different drivers led the model with only Tony Stewart and Matt Kenseth repeating. Greg Biffle's high ratings in his minor league seasons suggest what I already suspected: that Biffle likely wasted some of his potentially best Cup years climbing the ladder. While his leading the 2001 ratings is strange, it's less strange when you realize that he posted a 5-2 record in his Busch starts against Jeff Burton, the leader of the Cup brigade that year. Stewart leading for both 2002 and 2005 is no surprise as those were clearly without question the two best seasons of his career, although in retrospect it feels like almost every other year he had after 2000 was something of a disappointment, and indeed that is reflected by the model as he only makes one other top five aside from those years: 2007, where he battled Kenseth for best of the rest in the non-Hendrick class. Since head-to-head finishes essentially makes this a consistency-based model, I'm not surprised Kenseth led for 2003, especially since all his teammates failed to finish in the top ten in points. However, his 2007 lead is significantly more surprising. I guess it reflects that Gordon and Johnson were relatively evenly matched (although I think Johnson had the edge, and it's shocking that he didn't make the top five for that year at all, but I guess that's because Gordon beat him 18-13 in finishes that year. If I did the model based on laps led instead of finishes, I suspect Johnson would be leading for both 2007 and 2009. I definitely think Dale Earnhardt, Jr. and Mark Martin are being inflated for 2006 and 2008 because they had rookie teammates who had significantly worse rookie seasons than what they eventually accomplished, but 2009 may be more legit. Denny Hamlin did beat Kyle Busch 21-9, worse than he did in any other year, and worse than either Jimmie Johnson or Jeff Gordon did against Busch ever even in his earlier seasons (except for Johnson's 2007, which also resulted in a 21-9 defeat of Busch.) While I still think Johnson was the standout, I now think Hamlin wasn't so far behind as it seems like the Hamlin thrashed Busch and Logano while the Hendrick drivers seemed more even. That year is closer than I thought it was.

Although I don't really think 2004, 2010, and 2012 are Johnson's three best seasons I do understand why those are his best seasons in the model. In 2004, he benefited from Brian Vickers's first season and Terry Labonte's last both being pretty awful. In 2010, he won the title while all three of his teammates went winless and Mark Martin and Dale Earnhardt, Jr. in particular had awful seasons. In 2012, he benefited from Gordon's sudden decline. I would say he was probably the best driver all three of those years but not necessarily only those years. 2011 is a mess that I could have seen being led by any of a number of drivers, but seeing Kyle Busch atop that list is no surprise to me as it was one of his best early seasons. Although it seemed to me for a while around this period that the next era was going to be dominated by Kyle Busch, Brad Keselowski, and Joey Logano, that didn't really end up happening. Busch was always good but he tended to be closer to the back of the top five than leading it: for all his dominance, he really didn't have many consistent seasons and it seems like even those were inflated by the Toyota dominance, much like how rarely Martin Truex, Jr. appeared on the list in recent years (although I do find it interesting that both Kevin Harvick and Truex showed up a lot even before their dominant late-2010s period, indicating they were doing great stuff in inferior cars before they got their big chance, even if it often attracted little notice... Harvick did so often while Truex did so more intermittently but often enough that maybe his recent championship run shouldn't have been as surprising in retrospect.) But ignoring the inflated 2014 for Cole Whitt (due to his dominance over Alex Bowman before he was good), 2016 for Dale Earnhardt, Jr. (due to drawing a washed-up Kahne and Chase Elliott as a rookie), 2020 for Kurt Busch (due to a washed up Matt Kenseth), and 2021 for Chris Buescher (due to a washed up Ryan Newman), the story of the post-2012 period has been entirely about three drivers.

Throughout the late 2010s, especially in 2018, the NBC booth was constantly hyping NASCAR's "Big Three" because Kevin Harvick, Kyle Busch, and Martin Truex, Jr. were winning most of the races, but apparently they got the Big Three wrong. Judging by my model, if there was a Big Three (in terms of talent) it was clearly Chase Elliott, Harvick, and Kyle Larson. That triumvirate was the only group of drivers to sweep the top three positions two years in a row, in the exact same order both times with Elliott leading Harvick and Larson, although as I said if I removed all the Xfinity races from the model, Harvick would overtake Elliott in 2018 and Elliott would return the favor in 2019. Truex and Busch seem to have rarely ever been in the same league as Harvick. Since Busch's debut in 2005, he has only beaten Harvick four times in the ratings in 17 years (2008, 2009, 2011, and 2016, and Harvick beats Busch even in that year if I throw out the Xfinity races for both.) Anyone arguing for Busch over Harvick as the driver of the 2010s is wrong. It seems even relative to SHR, JGR and Toyota has had an equipment advantage, at least since 2015, which is the only reason it looked like Busch and Truex were in the same league. After the mid-2010s period, which Harvick dominated entirely by himself (even apparently including his last RCR year), the real standouts and rivals to Harvick however were clearly Elliott and Larson. This wasn't really noticed at the time because they were both driving underpowered Chevies in what was a very awful period for Chevy, although I certainly did notice the speed that Elliott and Larson had in 2018 when none of the other Chevies had that kind of speed and those two drivers could occasionally lead all the other cars of their make by as much as 10 or 20 seconds at times. It wasn't enough to make up their deficit to the SHR and JGR cars, but I certainly noticed it, and I'm actually not surprised to see them come out over Busch and Truex, even though Larson went winless in 2018. Since 2017, two of those three drivers have still placed in the top three every single year and they seem to generally remain the best drivers in the sport. Even though Hamlin clearly had the second best season, it's clear that the JGR cars were way faster than the SHR cars, and Harvick was apparently still putting up better performances for the cars he had. Considering Elliott usually beat Larson in previous seasons, it's surprising that Larson ended up with ten wins to Elliott's mere two (which dropped Elliott to an inexplicably low eighth in the ratings), but they certainly ran closer than that implies. Once Larson and Elliott had faster cars, I guess it's no surprise that they'd start dominating. The bigger surprise is how weak the Penske drivers are. Brad Keselowski never even makes a single list while Joey Logano only does so twice. Maybe they actually have had the fastest cars over the past several years (although not in 2021...)

While I certainly don't agree with every result my model produces, I still like it overall, particularly when evaluating 21st century drivers and seasons (although it is clearly way off when evaluating drivers who did not have teammates for most of their careers.) Although there are many bizarre outliers that usually resulted from drivers getting lucky to draw teammates that had inexplicably bad seasons, that is something I could eventually correct for either by simply removing seasons before or after a certain age, counting Cup starts only, splitting drivers' careers before or after injuries, and the like. What I really like is that the model does seem to correctly presage rising drivers' future success, as apparently mediocre-seeming years like Kevin Harvick's 2004 and 2005 (and even this year), Mark Martin's 2008, Martin Truex, Jr.'s 2008 and 2011, and Chase Elliott and Kyle Larson's 2017 and 2018, did seem to do a good job at predicting what those drivers would do in future seasons. The top fives tend to remain shockingly consistent lately from year to year with the exception of drivers who get the good fortune of drawing rookie or washed-up teammates. I now think it would be better if I did the model based on laps led instead of based on finishes, since I do tend to think dominance predicts the future better than consistency. When both Elliott and Larson were struggling in under-powered Chevies in 2017 and 2018, Larson was consistently leading and dominating more, which may explain in part why he was faster this season. I am sort of reminded about what happened in IndyCar when Josef Newgarden and Simon Pagenaud both arrived at Penske in 2016 and 2017. Both had impressed for second-rate teams but Pagenaud, like Elliott, was more consistent, while Larson, like Newgarden, was more dominant, and then Newgarden clearly took over control of the team from Pagenaud, the defending champion, almost instantly. Now I still think Elliott is way closer to Larson than Pagenaud ever was to Newgarden, but I do see a parallel there. Having said that, every driver does also have off seasons (Larson only ranked 9th in his last full-time season in 2019), so it's possible - perhaps even likely - that Elliott will revert to the norm and soon. What should really scare Larson's competition though is that for all his dominance, this wasn't Larson's highest-rated season, but actually trails his 2017 and 2018 performances. As implausible as it seems, he might be able to improve upon this (more likely, Alex Bowman and William Byron's ratings will continue to rise to lift him up in retrospect, or Larson was simply benefiting from a washed-up Jamie McMurray.)

My next big project will be to update my open wheel driver ratings to include the entirety of Formula One history (and maybe even some of the pre-F1 Grand Prix seasons) and IndyCar seasons as much as I can. Just like I did with this model, I want to include all the big names from every era of the sport, but I suspect that there will be some issues there too, because while multi-car teams have certainly been more common in open wheel racing than in NASCAR historically, few series had them as regularly and consistently as Formula One prior to the 21st century. However, limiting my previous model to simply drivers from the 2000s as well as their teammates was too limiting and I think it was the main reason so many of the '90s drivers were underrated. I'd like to see if that changes when I make sure to collect complete teammate records for all drivers in the model, which I did not do last time. I may not do this immediately (although I've already entered the 2021 data for all the races this year until last weekend's Formula One race) but I'd like to try to finish this before the end of the season, perhaps even before the end of the year. I've also thought about potentially doing another top 100 drivers list like I did in 2015, only this time it would be a lot more educated and informed but I suspect I won't have the time. Over the next few weeks, I may also provide complete lists of stock car drivers' ratings from best to worst for each season in a series of posts, although I realized I promised to do that for the open wheel model and didn't quite do that yet. Although there are many things about these models that I have to take with a grain of salt, they certainly give me a good headstart in preparing to rank the 1000 greatest drivers in racing history, now that my typing book is finished. I hope to be more active again, but I'll probably fail hard as usual.

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.

---