There are a lot of things people say in NASCAR that have been said for so long that nobody really thinks about where they came from – or if they’re true.
As you get older, you get more and more sensitive to generalizations people make about getting old. I can’t tell you how many times I’ve heard someone says that drivers get slower (or don’t win as much) as they get older. I started wondering if that was actually true.
The thing about data is that you have to think about what it means. For example, what does ‘slow down’ mean? Qualifying speed? Well, you don’t get points for qualifying first – we’re really talking about being successful at racing.
What racers really care about is winning – and if you can’t win, at least running consistently well. So I decided a good measure of “success” would be wins, top fives, top tens and rank at the end of the year. I used data from racing-reference.info to consider career statistics for drivers and look for patterns.
Given that time is finite, I picked some representative drivers. I wanted drivers with long careers, so I had data points at either end of the age spectrum. I wanted champions, so Johnson, Gordon, Kenseth, Stewart were obvious. I added Dale Jarrett because he had a long career and – hey, I just like Dale Jarrett, OK? I also wanted some youth, so I added Kyle Busch, Kasey Kahne and Carl Edwards. I added Ryan Newman because he made a big splash the first couple years of his career and Jeff Burton because (in addition to a long career), he’s had a really interesting career. Finally, I added Mark Martin because – well – he’s been around forever.
Let’s look at Jarrett’s data first. The number of wins is in blue, top fives in red and top tens in green. I’ve indicated the year he won the championship with an arrow (which represents his best year in terms of top fives and tens, but not his best year in terms of number of wins. Also worth noting: He didn’t run full seasons his first three and his last two years. But even if you ignore those, there’s a pretty clear pattern of improvement and then tailing off.
Tony Stewart didn’t start Cup until he was 28, though, so you might expect him to have a similar profile to Jarrett.
Not at all! Stewart has been remarkably consistent over his Sprint Cup career. The two arrows show Stewart’s first and last championship years. The third was at age 34. Stewart doesn’t seem to have any peaks or valley, even when he changed teams. Jimmie Johnson’s numbers look very similar to Stewart’s. Kenseth’s is a little more up-and-down, but again, there aren’t noticeable peaks. (I know, I should have done Kurt Busch to see whether this is a mark of a champion!) I’ll get there…
Each of the other drivers have similarly different (?) graphs. Newman peaked at age 25 and he’s been up and down since then (with a slight upward trend since he joined SHR.) Edwards is very up-and-down. Martin’s is a little complicated because he had a couple years he didn’t drive all the races. Even so, there’s a broad peak from about age 30 to age 40 and everything after that is pretty oscillatory.
Here’s one of the complicating factors, though. Look at the data for Jeff Burton – he’s got two peaks and they are very clearly divided by the year where he switched from Roush to Childress. Similarly, Kasey Kahne’s numbers show a big upswing when he started driving for HMS.
And then there’s Jeff Gordon… I spent a lot of time trying to overlay life events on his graph – could you correlate performance to marriage, divorce, birth of children. In short, no you can’t. Gordon shows a broad peak from age 22 to age 28-29 and it’s been a little disappointing since then. There is the occasional very good year (six wins in 2007), but it’s mostly down from his heyday. Since he has the same equipment as Jimmie Johnson, it’s hard to make an argument that it has to do with the company falling down on technology. Sorry – wish I had better news there.
Umm… So… You Gonna Tell Us Anything Useful?
This, of course, raises the important question of “How are you going to come to any type of conclusion if everyone’s graphs look very different?”
Which is a very good question. I played around with the data and I finally came up with an idea. What if I take a bunch of drivers and simply add together how many wins were had by drivers that were aged 21, 22, …? I did that.
Then I realized that was stupid. Why? Because that’s really measuring how many drivers of each age there were. It looked very Gaussian and that’s what made me smack my forehead and realize I was being stupid. What I had to do was find the average number of wins: How many wins by 28-year olds divided by the number of drivers in my sample who were 28 years old. Now this has some promise…
There’s still a lot of scatter in the data (I haven’t had time to increase the numbers of drivers), but two clear trends emerge. The top graph is a plot of the average wins per driver vs. driver age. Jeff Gordon is messing things up a little with his 13 wins in 1998. He won just about 40 percent of all the races that year. But still – if you look at the winner for a given race, the odds suggest that he or she will be younger rather than older. The outlier at age 50. That is Mr. Martin. He won five races that year and I only have data for one other 50-year old driver.
The lower graph shows the number of top tens (blue) and top fives (red). Those numbers are both surprisingly flat until you hit about 40 years old, at which point they start going down.
Now you have to be careful – this tells you the behavior of a group. It doesn’t tell you the behavior of any one particular driver. When doctors tell you things like survival rates for cancer, they don’t tell you you have a 50% chance of being cured. They tell you that, out of people like you, 50 percent of them are cured. Your individual rate could be 10 percent or 90 percent. Of course, that’s not something we can measure.
Just like we can’t measure luck – if you get taken out by an inexperienced driver, that really doesn’t have anything to do with your age.
Now there’s the big question. It is a fact that our reaction rates slow down as we age. But we also gain experience, which may allow us to make better decisions. The relatively flat profile of a driver like Stewart doesn’t seem to indicate any decline over the years. The Stewart of his first championship is just as good as the Stewart of his second or his third.
As Silly Season progresses, teams are going to be more likely to look for a younger driver (like Kyle Larson) than take a chance on a veteran like Bobby Labonte. The older you get, the more you have to prove yourself.
NOTE: My friend Josh Browne pointed me to another stats blog that covered some of the same ideas – this is a great blog and should be of interest to racing geeks everywhere!