Last Year's Rolex

by Sean Wrona

I realize three days after the finish of the 2015 Rolex 24 at Daytona, more people would likely be interested in reviewing the current race than the previous one and I do intend to do that eventually. However, I spent most of my nights last week on this to try to get it done before the start of this year's event and I wasn't quite able to manage it. Now that I've compiled all this data, I certainly don't want to put it to waste.

I initially intended to calculate average running position only for laps each driver was on the lead lap, but for sports car racing, I came across a problem. When multiple drivers drive a car especially over races that last many hours, many drivers will already be laps down when they first enter the car. In order to correct for this, I was forced to use average running position including all cars running on the track, but obviously this hurts several drivers who entered the car for the first time when they were far out of contention even if they were among the fastest drivers in their class (most notably Sam Bird). While there is still some value in calculating average running position in this way, in situations when drivers entered the car for the first time when it was far out of contention, this statistic must be taken with a grain of salt. As I mentioned before in The Racermetrics Manifesto, I calculated speed based on laps listed as exclusively green for which the lap time was within 10% of the fastest lap time in that class...this was pretty self-explanatory. I threw out all drivers who did not have ten competitive laps in the race (where a competitive lap is defined as within 10% of the race fastest lap). I compiled a list of every pass per position in the race for all four classes to calculate my passing statistic (which was easily the most time-consuming part of this analysis). I counted all laps where drivers changed position and had comparable lap times throughout the lap. This can be a judgment call at times but generally, the only on-track passes for position I did not count were those where there was a significant lap time difference and a vast majority of the lap-time difference occurred only in one sector (yes, I checked every single pass for position for this). If somebody had an unusually slow sector time, that indicates that driver likely had a spinout or off-track excursion and that shouldn't necessarily be counted as a 'pass' since in a situation like that the drivers passing didn't actually do anything except avoid a wreck (and there is some merit in that, don't get me wrong). I did not count any passes of lapped cars or cars in any classes. As I mentioned in the earlier column, I awarded a number of points equal to the number of competitive drivers for passing the fastest car and a single point for passing the slowest car, with a loss of one point for being passed by the fastest car, and a loss of a number of points equivalent to the number of competitive drivers for being passed by the slowest competitive car. (If a driver was passed by a driver who did not have ten competitive laps - and that did happen - I deducted a point more than I did for being passed by the slowest car). The whole point of this is to distinguish passes the drivers made themselves versus those they inherited. If there was a large lap time difference and it was spread over multiple sectors (i.e. the passing car was significantly faster everywhere) I counted it as a pass. Finally, in order to rank natural peak (my measure of dominance) I ranked according to (in order): highest position obtained via a green-flag pass for position (with drivers who retain that natural peak position for the most consecutive laps ranked higher), then highest initial position for drivers who made a green-flag pass for position lower than their initial position, then highest initial position for drivers who did not make a green-flag pass for position. To make this a bit clearer, it is necessary to provide examples. Christian Fittipaldi led upon entering the car for the first time in the 2014 Rolex but did not make a single green-flag pass for position in the entire race. He led 63 laps in the race after inheriting the lead when he entered the car, so he had a higher natural peak than all the drivers who made passes for 2nd in the race, but is behind all drivers who made a green-flag pass for the lead and retained it, even if they led for fewer laps than Fittipaldi did. I do make an exception for polesitters who retain the lead on the track, since winning the pole is something that the driver does. Since Alex Gurney won the pole and led the first 21 laps of the race (even though he made no green-flag passes in the race) I gave him credit for a natural lead, but Fittipaldi I did not. I would not have given Gurney credit for leading naturally and would have placed him behind Fittipaldi if one of his co-drivers besides him had qualified the car. To further explain how I broke the ties here, Justin Wilson made a pass for 2nd on the track and retained it for 7 laps, Richard Westbrook entered the car in 2nd place but made his highest pass for position in the race for 7th, while Wayne Taylor entered the car in 2nd place and made no green-flag passes for position. Using this statistic, I ranked Wilson highest, Westbrook next, and Taylor lowest, which seems reasonable. While I do count positions obtained upon entering the car, the tiebreaker always goes to the drivers who actually made passes for that particular position rather than those who just inherited it. For drivers who led in a particular class, I gave credit for laps they led on multiple pit cycles if they happened to lead before and after a green-flag pit cycle. What I'm attempting to measure here is control of the race, and you do not necessarily have to lead consecutive laps to maintain control of the race. However, it would have taken considerably more effort to do this for every position, so I only counted laps maintaining a position for class leaders. A driver who made a green-flag pass for the lead multiple times would get credit for all the laps combined from each stint since I am attempting to measure dominance here and one can argue that making multiple passes for the lead on track in fact one of the best measures of dominance.

The Prototype class epitomized one of the central fallacies of racing analysis, particularly assuming that the winning team had the best performance in the race solely because they won. Even though the #5 Action Express Racing Chevrolet Corvette of João Barbosa, Sébastien Bourdais, and Christian Fittipaldi won the race and even though they also led the most laps, when you consider how they led, not just that they led, one is forced to conclude that Wayne Taylor Racing's #10 car drivers Max Angelelli and Jordan Taylor outdrove them even though they did not win the race. Taylor and Angelelli had the fastest two average lap speeds in that class (Taylor's 101.67 and Angelelli's 101.75 to Barbosa's and Bourdais's 101.82). That may not seem like a significant difference to you but when you consider that Taylor ran 176 competitive laps and Barbosa ran 201, Taylor's 0.15 second advantage approached 30 seconds over the entire race, and that is pretty substantial. Furthermore, five drivers led naturally in Prototype: Jordan Taylor, Max Angelelli, Alex Gurney (from pole but not after that), Scott Dixon, and John Martin. Notice that both the #10 drivers made such a pass (although Ricky Taylor did not) and none of the #5 car drivers did despite winning the race. Fittipaldi did enter the race as the leader, but he inherited the lead when Barbosa pitted. Scott Dixon led Barbosa on track before that cycle of pit stops, and Dixon pitted and was replaced by Tony Kanaan, while Barbosa pitted and was replaced by Fittipaldi. Fittipaldi then managed to beat Kanaan out of the pits, hence he does not get credit for an on-track pass for position, nor does Barbosa (since he only took the lead in the first place because Dixon pitted). The entire rest of the race continued like that with the #5 only taking the lead in the pits, while both Taylor and Angelelli made on-track passes for the lead, with Angelelli passing Bourdais and Taylor making three natural passes for the lead (of Jamie McMurray, Barbosa, and Martin). The #5 only ultimately took over the lead because Taylor pitted on lap 633 ten laps after he had previously pitted under a caution while Bourdais did not choose to pit, then the #10 was unable to pass the #5. Action Express Racing won on strategy and that's certainly part of the game but it doesn't mean their drivers ran the best. The #5's two lead drivers got passed for the lead by the #10's two lead drivers but the #5 never made an on-track pass for position despite leading the most laps. While I weight dominance heavily myself in evaluating driver strength (and much more than either NASCAR or IMSA do in their points systems) when evaluating drivers, I certainly weigh drivers dominating based on their own efforts more than drivers dominating because they inherited a position on a pit cycle. Don't get me wrong. There is skill to that. Dixon beat the #10 car in this year's race for largely that reason, but it's still a lot easier to make a pass on a pit cycle than on the track, and that makes on-track passes more valuable because of their relative scarcity. Honestly however, I think the #10 likely would have won both last year and this year with better strategy. It's worth noting that Action Express Racing is currently employing legendary NASCAR crew chief and ex-NASCAR director Gary Nelson, who certainly knows a thing or two about strategy calls and engineering setups, while Ganassi has consistently had brilliant strategy in the Rolex Series. Judging by the #10 team's questionable strategy in this year's event (I personally thought they should have left Angelelli in the car for the rest of the race this year once Ganassi announced that Dixon would be running the last four hours; it was too big a risk to replace Angelelli with Jordan Taylor when Ganassi would not be making a driver change and the #10 had less than a ten second lead) it seems that their strategy calls in general are not as consistent as Ganassi's and Action Express's. However, that doesn't mean that their drivers weren't more impressive than the #5's in last year's event. Does that necessarily mean that I think Jordan Taylor was the best Prototype driver overall in last year's race? No. The Chevrolet Corvette DPs did have a gigantic advantage over anything else in the Prototype class, and I am possibly more impressed by Dixon (who was the only driver to lead the Prototype class naturally in a Ford EcoBoost, or anything other than a Corvette DP for that matter) or Olivier Pla (who despite driving a P2 prototype, which were much slower in 2014 than the DPs) still managed to be 7th in average speed and more than 6 tenths of a second faster than his teammate Gustavo Yacamán, the next fastest Le Mans-style prototype in the race. That is an astounding difference between teammates, and even more astounding considering that even Yacamán was faster than all the Extreme Speed P2s in the race despite the great number of talents that team had entered in the race (Simon Pagenaud, Ryan Dalziel, David Brabham, and Scott Sharp). When you consider the equipment difference, I might have to give it to Pla over Taylor because Taylor did not dominate the other DPs by nearly as much as Pla dominated the other P2s. However, as far as the drivers who were actually driving cars competitive for the overall win, Jordan Taylor was easily the best driver in Prototype, even though he didn't win. I did not take equipment strength into consideration when I calculated my driver ratings and only calculated based on performance alone, as I haven't come up with a great objective measure for equipment strength yet, but I will eventually do this, and if I did do this, I would almost certainly have to choose Pla as the top driver in the Prototype division.

The other three classes ended up being more run-of-the-mill in the 2014 Rolex. The top drivers according to my ratings system were all class winners except for Taylor: Colin Braun in Prototype Challenge, by quite a lot actually, Nick Tandy in GT Le Mans, and Jeff Segal in GT Daytona. Because GT Daytona had so many more drivers than the other classes (a total of 113 with ten or more competitive laps), passing was far more frequent so there were very noteworthy distinctions between drivers in the passing and natural peak categories. Prototype Challenge by contrast was extremely boring. Since there were only nine cars entered in the class and the class is completely spec with only one type of chassis and engine, there were only eighteen passes for position in the entire race and only two after lap 131. Because many drivers in that class neither made a pass or were passed in that race, there was a massive tiebreaker in the passing statistic for those with an overall passing score of 0 (while all four classes had a number of drivers that this applied to, it was by far worse in the PC class because of the lack of passing). Passing data obviously become more and more reliable when there are many more passes, but when very few drivers made passes at all, I would have to say my Prototype Challenge rankings are the least accurate of the bunch, especially when you consider that Sam Bird had the fastest average lap time but appears very low because the car had already been effectively eliminated from the race before he entered it. The Sam Bird anomaly convinces me that perhaps I should completely ignore average running position and finish entirely for sports car racing and only focus on speed and passing, because on a sports car team one driver shouldn't be blamed for what the previous drivers in the car did or did not do. I hoped that I could use this formula as a kind of one size fits all formula that I could apply to multiple different series, and I may still do that, but I really should have known better than to concoct a one size fits all formula when each form of motorsports have some different criteria that need to be factored into the results. I shouldn't rate drivers on single-driver teams and drivers on multi-driver teams the same way. I think I believe that pace (raw speed) and on-track passing and natural peak are likely the best predictors for what will happen in the next race, so a much simpler formula could just include these two factors. I am definitely convinced by this that I was seriously misguided to rate speed at only 10%. The reason I did that was because I did not want to reward drivers solely for being in faster cars, but obviously all the other categories also reward drivers solely for being in faster cars, but at least speed includes what the driver does to a greater degree than perhaps something like average running position which can be greatly affected by strategy (or just luck or other drivers' mistakes when I am including drivers not on the lead lap, but if I don't include drivers not on the lead lap, then some drivers could not be rated at all). While I still think average running position is pretty accurate for single-driver series, it isn't here because drivers who remain on the lead lap for the entire race get a massive advantage (since there are so few cars on the lead lap at the end of the race), drivers who encounter trouble early get a massive disadvantage, and so on. I'm actually growing increasingly convinced to measure solely speed, passing, and natural peak and nothing else at least for sports car series (should a weaker driver be rewarded for getting a good finish thanks to superior teammates, or a stronger driver be rewarded for getting a bad finish thanks to weaker teammates?), but I may continue to calculate all five for single-driver teams. Honestly, my opinions are evolving for what most impresses me. I used to merely say dominance impressed me more than consistency (at least when comparing a dominant but inconsistent driver to a consistent but not dominant driver), but now I would say driver-induced dominance is what truly impresses me while team-induced dominance really doesn't when evaluating drivers (with some obvious exceptions: Scott Dixon certainly did contribute to the winning move in THIS year's Rolex even though he passed Jordan Taylor off-track, but I would be still more impressed if he had passed Taylor on-track). The #5 may have led the most laps in Prototype but the #10's drivers were more impressive last year. I firmly believe that. I haven't done 2015 yet, but I suspect I will find that the Ganassi #02 car was more dominant than the #10 car this year (or at least Dixon was more dominant than anyone on the #10 car).

Obviously the results were too car-centric here when drivers who drove slow cars were right at the bottom together (the four drivers of the #0 DeltaWing car in Prototype had the bottom four race scores) for instance, but this does nevertheless allow for some interesting comparisons, especially when drivers were some distance off from their other teammates. Team owners who tend to hire better drivers than themselves are as expected quite a few positions behind their teammates, but even among the pros, there can be large distinctions. Bourdais was the best driver in the #5 in 3rd place with a race score of 92.8 out of a maximum possible score of 100, with Barbosa in 4th at 86.5 and Fittipaldi in 10th at 81.6. These are pretty significant differences, largely due to Fittipaldi not making a green-flag pass in the race and also being passed in the race by Dixon, which placed him behind the gaggle of drivers who scored 0 passing points and near the bottom in passing overall despite having one of the best cars. This might not be fair since Barbosa, not Fittipaldi, started the race and in a road course endurance event, it is much easier to make passes early on when there are many cars on the lead lap than later when there are fewer, and perhaps it's unfair to penalize Fittipaldi when he did enter the car as the leader (since you can't gain any positions as the leader, but he did do a nice enough job of maintaining them to lead all the drivers in average running position). Similarly, Burt Frisselle went on a passing clinic early in the 2014 race and led in passing, but Taylor and Angelelli were arguably more impressive there as well. Even though Angelelli was 2nd in passing and Taylor was 3rd, they were very close to Frisselle yet Frisselle started the race and they didn't (Ricky Taylor started the race in the #10 car, so he and Barbosa and Burt Frisselle had many more opportunities to make passes than Angelelli and Jordan Taylor did because they arrived in the cars later in the race when there were fewer cars on the lead lap, but Angelelli and Taylor still nearly outshone everyone else in this category). There are a lot of things to quibble about with all these statistics, and I now think for this particular series, average running position and finish may be almost worthless to include (especially given the endurance aspect of this particular race). It may simply be better to not attempt to combine these diverse categories into one messy whole. What I will say is that leading any of these categories is a pretty big deal and a strong indication of future results (especially if you are leading a category that depends on what you did, rather than something like average running position or finish that can depend on what your teammates did).

Below are the top ten drivers in each category in each class (I have omitted the finishing positions, which you can find at http://www.race-database.com/results/results.php?year=2014&race=1&series_id=14.

Below are the overall rankings for each class with each driver's score in each category (based on the percentage of other drivers beaten in that category, with a maximum of 30 points for average running position, 20 points for finish, 10 for speed, 20 for passing, and 20 for natural peak, but obviously I am going to rethink that since I now think I have seriously undervalued speed, which one could easily argue is the most important):

Even though the 2015 Rolex just ended a few days ago, I'm probably going to wait a few more weeks before finishing the analysis on that, particularly since I think I'll likely want to change the weights I'll give to each category based on this analysis and while the idea of a one-size fits all driver rating that could apply to a bunch of different series is appealing, it's not very realistic. It would make more sense to not attempt to add the categories together and merely spotlight the drivers who were overlooked because they were very good in speed or passing relative to their teammates even if the team itself didn't do well overall and I may do that in the future, but I wasn't going to waste all this good data and keep it to myself. My next few columns will likely be much less strenuous as I will be focusing on my upcoming "How the Races were Won" series which determines whether races were won naturally via an on-track pass, via an off-track pass, or due to sheer luck (incidentally). Even though I imagine many more people would be more interested in discussing the just-finished Rolex after this one, this analysis was painstakingly time-consuming and I can't do one of these for an endurance race this often (although I probably will do one for every Formula One and IndyCar race, and I'd love to do it for NASCAR Sprint Cup too, but NASCAR does not provide positions and lap times for each driver for each lap unlike most other sanctioning bodies. Trying to get this column done before the Rolex and then failing to do so and releasing it almost two weeks later was not the most efficient way to launch a site, but I still think there's some cool stuff in here that is worth looking through if you care about American sports car racing. I did calculate the average lap times for this year's Rolex already and I may put those up as an article by themselves in the next week or so, but I don't intend to go this intensive with regard to the 2015 Rolex in the very near future.

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. You may contact him at sean@racermetrics.com.

---