Rules changes in NASCAR are tricky. There’s really no way to test them before implementing them. You’re relying on the judgement and experience of the NASCAR team, with input from race teams, drivers, Goodyear and tracks.
How to Test Rules Packages?
Robert Gauthier (@yetticrg) tweeted a question to me (You can, too: @drdiandra). I answered it on twitter, then had that immediate panic of ‘Wait! Does the data really support my instinct?’
Spoiler: It does.
So here’s Robert’s question
It has been advocated to run all star package at a track one race and then the current package at the same track next year. Are tracks that consistent to make it a valid one to one comparison?
The idea is that you could use the All-Star race to test a rules package for that particular track. It’s a nice idea because out of necessity, you’d have to move the All-Star race around each year. The issue with that idea is that this would be a years-long experiment. And if there’s one thing people want from NASCAR, it’s speed.
But the question of consistency is relevant in another context as well: NASCAR runs 36 points races at 24 tracks, which means that we visit 12 tracks twice a year.
|Bristol||April 16||August 18|
|Daytona||February 18||July 7|
|Dover||May 6||October 7|
|Texas Motor Speedway||April 8||November 4|
|Kansas||May 12||October 21|
|Las Vegas||March 3||September 16|
|Martinsville||March 26||October 28|
|Michigan||June 10||July 12|
|Phoenix||March 11||November 11|
|Pocono||June 3||July 29|
|Richmond||April 21||September 22|
|Talladega||April 29||October 14|
Nota Bene: I was surprised. We only visit three mile-and-a-half tracks twice. With Fall Charlotte going Roval, there are only ten mile-and-a-half tracks on the schedule.
So let’s look at the question: Can you test one rules package at one race and then compare it to another package at the second race of that year? Or even the next year?
The scientific method you learned in school, which is illustrated at the top of the blog, is wrong. It’s more like this:
The one thing you probably remember, thought, is that you had to pick one variable and then hold everything else constant. You’d plant beans and put one pot in the closet, one in the brightest window, one in the guest room where the shades are always down… and then you’d compare the effect of light on the plant growth.
These type of experiments are lovely idealizations, but possible only in laboratories, where you can control a lot of things. Any scientist or engineer who works with natural phenomena (or people) can only control limited things. Think about people who research earthquakes, supernovae, or gravitational waves. You can’t just order those up when you’ve got everything ready.
And, unlike scientists, NASCAR has to run their experiments in front of people. You can use tests to eliminate options that don’t work, but testing doesn’t replicate the conditions of a real race. If you run a stupid experiment in your lab, no one but you and a few other people know. If NASCAR screws up, everyone knows.
So is it possible to run controlled experiments of the type Robert and I have suggested? Let’s look.
NASCAR is a sanctioning body. They make the rules. But race cars are complicated and they must deal with a huge number of variables: spoiler, splitter, sway bar, track bar, four springs, four shocks, body shape…
And then each one of those variables has multiple possible values. Look at the spoiler.
And NASCAR can control each one of those. But there is one big thing that even NASCAR can’t control: the weather.
Temperature is one of the most significant variables because air temperature affects track temperature and track temperature has a huge impact on grip. And racing is all about grip. Look at the differences between night races and day races at the same track.
Let’s take Texas Motor Speedway as an example. The NOAA (National Oceanic and Atmospheric administration) website lets you look at average monthly temperatures as a function of time. They go back to 1930, but I thought going back to 1990 would be just fine. Let’s look first at April.
And yes, I know the track is closer to Fort Worth, but NOAA doesn’t have data for Fort Worth.
The average temperature in April (over these years) is 67.0°F. I plotted November on the same scale (note the suppressed zero on both) for ease of comparison
November in Dallas is, on average, nine degrees cooler (58°F). There’s not only a significant difference from April to November, there’s also a significant year-to-year variation. It was 50.9°F in 2000 and 60.6°F in 2001.
In the interest of looking into whether we can test something at Texas in April and then use it in November, let’s plot the difference in temperature from one spring to the next fall:
The difference can be as much as sixteen degrees.
Daytona goes the other way. The average July temperature (82.0F) is much warmer than February (61.3 F)
Temperatures range from 53.4°F to 68.9°F in February, so year-to-year poses a challenge for comparison.
While the temperature variation is much smaller from one July to the next (3.7 degrees), the temperature differential from February to July within a single year can be 30 degrees.
Ah, but you’re thinking… Michigan and Pocono. Because of their locations, those two tracks host races in June and July. I’m using the data for Detroit (Michigan) and Allentown (Pocono) and I’ve plotted only the differences from June to July.
Detroit varies from 65.4°F to 74.0°F in June and from 68.8°F to 79.3°F in July. The differences from June to July can be almost nine degrees.
We range from 67.8°F for the coldest June in Pocono to 78.7°F for the warmest July in that time period. Again the temperature differentials for June to July can be up to eight degrees.
These are the two most replicable tracks, weather-wise; however, there’s such a short time between them that anything you learned in June probably couldn’t be implemented for July just because there wouldn’t be enough time.
Assuming the All-Star race would be run the weekend before or after the already-scheduled race, it’s possible that Robert’s scheme of testing something for during the All-Star Race for the next year might work. The downside is that results from Michigan are likely transferable only to Fontana and results from Pocono are, well, results applicable to Pocono.
You don’t need to look at graphs to know that rain and snow are utterly unpredictable. Both can significantly change the race track.
Perhaps more important than the amount of rain is when the rain comes. If it comes the night before the race and the track is green at racetime, that’s a different situation than if it rains on Friday and the track has all day Saturday to rubber up.
It’s not just the rain that changes the track. Drying the track also changes it. Drivers have commented that the Air Titan can be used (and has been used) to condition the track in way that weren’t possible with jet dryers.
Tracks age. We normally think of aging in terms decades, but there are little changes every day, which add up to meaningful changes over months. Asphalt wears, making tracks rougher and more abrasive. The ground settles, leading to bumps and dips in the track surface.
All of these change require human intervention to repair. The ‘silicone injections’ at Texas to fix sagging spots in the track, the mid-race track patch at Daytona and wholesale changes like resurfacing all change the nature of the track.
Most tracks experience the freeze-thaw cycle, which really does a number on asphalt. Asphalt (and concrete to a lesser extent) are porous. Water within the pores of the asphalt freezes. Water is unique liquid: one of the very few things that expand when they get cold. The expansion pushes on the asphalt, creating stresses and eventually cracks. The more freeze-thaw cycles, the more stresses on the track.
And let’s not forgot the occasional natural disaster. Hurricanes impacted Daytona in 2009 and 2016, flooding the track and submerging the racing surface.
Most race teams know more about how to make their cars fast now than they did in January.
Scientists can sometimes come up with creative ways to account for variables they can’t control: for example, grouping people in medical studies by age, sex and race/ethnicity.
The problem is that this only works when you have large numbers of datapoints. With 36 races each year, that’s not a luxury NASCAR enjoys. The smaller the number of data points you have, the less confidence in the accuracy of the predictions you make from them.
The Human Element
NASCAR has no control over the weather and only limited control over human beings. Another thing NASCAR can’t stop: Progress.
School’s Always in Session
NASCAR teams are constantly learning about the car. Wind tunnel and computational fluid dynamics simulations, seven-post rig tests, real-time race data and much more provide the engineers with a stream of data to analyze and learn from. Each practice teaches the team something new about the car. NASCAR can’t stop the teams from learning and there are many more engineers working for race teams than there are working for NASCAR itself.
The Biggest Variable
When a engineer looks at the car, the biggest variable he or she sees is sitting behind the wheel.
A winless driver nearing his hundredth start really wants not to have to hear people talk every week about when he’s finally going to win. He runs a race with no cautions in the last stage — which is good because he doesn’t finish as well in races with a lot of restarts. He wins the race. If there had been a lot of restarts, he may not have won. If he hadn’t been going for his first win, he might not have taken chances. If there had been a lot of restarts, he may not have won.
Just look at how many variables are in that one paragraph.
Whether there are a lot of restarts depends not just on our focus driver, but also on the other thirty-something drivers out there. The driver who’s been told he already doesn’t have a ride next year and is trying to prove something, the driver who had a fight with their wife last night, the driver who’s already in the playoffs and the driver who’s got two chances left to make the playoffs…
All these things not only make a difference, they’re not things we can predict, they’re not things we can measure and we certainly can’t compensate for them.
And that’s not even accounting for the driver who loses his temper and purposely crashes out another driver. Talk about unpredictable.
There’s no surefire way for NASCAR to test their race packages. They have to rely on the institutional, collected knowledge of the people working in the sport to make decisions, along with data from a limited number of carefully planned tests. Even if we could control track conditions, there’s no way to control the drivers — some of who have proven themselves to be much bigger variables than others.
It’s a fact of live sports that some events will be less thrilling than others. As much as NASCAR wants every race to have people on their seats throughout, that’s not going to happen. I fear it’s part of the ongoing compulsion people have developed that requires them to be constantly stimulated, whether it’s social media, eSports or gaming.
If you’d like to learn more about why this need to be constantly stimulated might be a really bad thing, I recommend this episode of a BBC show called ‘The Why Factor’, which about why it might be better for all of us if we were bored more often.