Fifty percent races won at Talladega were won by drivers qualifying from the first two rows. One-third of the winners started from the first row.
What does that tell us about where Joey Logano — or any driver who needs to win to stay in the playoffs — must qualify?
Winning vs. Qualifying Position
The histogram below shows how many winners (vertical axis) started at each qualifying position (horizontal axis) in the100 NASCAR Cup-level Talladega races to date.
- 14% of the winners started on the pole
- 34% of the winners started on the front row
- 50% of the winners started in the front two rows
- 90% of the winners started from 18th place or better
- No one has ever won starting from higher than 36th
- Only 5% of the winners have qualified 27th or higher
This would suggest that teams that must “win to stay in” should focus all their efforts on qualifying well.
But, of course, that’s not the whole story.
Qualifying Over Time
A large number of data points gives your stronger conclusions and nicer-looking graphs. We’ve built up a lot of race data for most tracks. But remember that NASCAR is always changing, and changes are often hidden when you reduce the data too much.
Let’s look at the qualifying positions of Talladega winners over time.
I’ve highlighted the top five qualifying positions in a red (well, pink) box. Pick a range of years and then compare how many points are in that box to how many are outside. The ratio is very different if you’re looking before about 1998 vs. if you look after 1998.
We can quantify this difference by looking at averages over different periods of time.
- The winner’s average starting position from 1969-1998 was 5.2.
- From 1998 to the present, it’s 12.1.
There’s not really a hard line: There’s nothing magical about 1998. The difference gets even more interesting when we look at average starting position by decade:
Before 2000, most of the winners came from the first three rows. More recently, winners have come from further back and being in the front row (or front two rows) isn’t as important.
The standard deviation is the amount of variance in a set of numbers.
- A small standard deviation means that the numbers are all close to each other.
- A large standard deviation means that there’s a lot of spread in the numbers.
- 68% of all the numbers within a set lie within one standard deviation of the average.
So here’s the data above, but plotted to include the standard deviation.
- In the 1970’s, 68% of the winners qualified between 1st and about 14th.
- The spread was even smaller in the 1980s: 68% of the winners qualified in the top 8.
- In later years, the spread has gotten even larger: winners are coming from further back in the field.
What’s really interesting here is that, in the 2000s, winners were actually less likely to come from the front row. In the 2010s, so far, winning from the pole makes you an outlier.
Why the big changes over time?
A lot has changed in the last two decades. Just to name the ones that come immediately to mind:
- Computational fluid dynamics and wind tunnels revolutionized the way teams built their car bodies.
- Remember when drivers had a dozen radio channels in their cars so they could strategize with other drivers?
- Tandem drafting changed the nature of racing so much that NASCAR mandated changes in the cooling system to discourage it.
- Some drivers adopted a strategy of hanging in the back to try to avoid the big wrecks
- Manufacturers mandated cooperation only with teams driving the same make.
There are many other changes, all of which are hard to quantify, but I think we can pin down at least one major change that’s made qualifying up front a little less important at Talladega.
DATASET: The following analysis is based on all Fall Talladega races from 1970-2018.
One thing I wanted to look at was the number of cars that didn’t finish the race, because we’ve seen a lot of times where front-running cars get taken out in a crash and I wondered if that had changed over time.
This is the percentage of the field that didn’t finish: I didn’t use absolute numbers of cars because the field size has changed over the years.
The DNF rate definitely decreased — with the exception of the wacko 2017 race, where 62.5% of the cars in the field didn’t see the checkers — but the decrease wasn’t as much as I had expected.
You can see the change a little more clearly if we break it down by decade.
Restrictor plates were instituted in 1988, so you might think that was the reasons for the decrease was fewer crashes.
You’d be wrong.
Reasons for DNFs
Over the 50 races I examined, NASCAR gave 42 different reasons for why cars didn’t finish a race. That’s what ‘s in the word cloud in the header.
They were a lot more specific back in the day, noting whether a failure was due to a valve, a valve spring or a crankshaft.
Some are more descriptive than others. In the 1994 fall race, the car driven by three-time ARCA champion Tim Steele is listed simply as “quit”.
I combined the 42 reasons (which accounted for 708 DNFs) into five groups.
I split out engine failures from mechanical failures. The last two categories barely contribute to the total and I won’t do anything further with them.
DNFs Due to Mechanical Issues
This graph shows the percentage of cars that start the race, but DNF due to mechanical failure. It is not the % of DNFs, but the percent of cars in the field that DNF due to mechanical failure.
In the 70’s and most of the 80’s, it was not unusual for a third of the cars to drop out of the race due to mechanical failure. There were four races where 50% of the cars DNF-ed due to the car failing.
We just don’t see that anymore. In the last 10 years, we haven’t even reached 10% of the cars DNF-ing due to mechanical failures. The majority of those failures are engines giving up, but even that has become much less common in recent years.
So where do the rest of the DNFs come from?
DNFs Due to Crashes
Again, this is the % of cars that started the race and didn’t finish due to being involved in a crash.
On average, the number of DNFs due to crashes has increased over time, even if you toss out 2017 as a fluke. While we tend to think of Talladega as a track with a lot of crashes, it hasn’t always been that way.
Why It Matters for Qualifying
The probability of exiting a race due to a mechanical failure has gone down, but probability of getting knocked out by a crash has gone up. That’s played havoc with the relationship between where you start and where you finish.
Look at how so many of the data points lie on a straight line for the highest finishers in 1978. (I drew the line to make it easier for you to see.) Most of the drivers at the later finishing positions (the ones off the line) didn’t finish the race.
In contrast, you could draw a straight line for the 2018 race, but the data points would be much further away from it. There’s less correlation between how well you start and how well you finish today.
We’ve changed from being knocked out by something within your control (i.e. your car) to something often outside of your control at Talladega: your ability to stay out of crashes.
If I’m a crew chief, I want to make sure we qualify decently — within the top 15 or so — but I’m not going to stress out over the difference between first or fourth. In the end, it probably won’t make a whole lot of difference.
I’m also going to have my spotter(s) keep an eye out for potential trouble spots to try to help him avoid a crash, which is a much bigger threat to remaining in the playoffs than not getting the pole.
Gratuitous Graph for Moody
This is all the DNF data from the discussion above put into one giant, colorful graph.
This graph actually has a purpose. You see, scientists love graphs because, if a picture is worth a thousand words, a graph is worth ten thousand words. Because we speak “graph”, if I were writing this for a scientific audience, I wouldn’t break all the data down, but just put this up there.
The more you learn how to understand graphs, the easier it is to understand the data for yourself — and not have to rely on someone else, who may not telling you what you need to know, or may be trying to slant the data to support their own position.