Was Phoenix a ‘Bad Race’?

A lot was written last week about the Phoenix race being a ‘bad race’ — but was it?

In comparison to what? The typical race at Phoenix? Our memories of races at Phoenix? Our expectations?

I argued before that it’s worth considering the numbers and the history in addition to our gut feelings. So let’s do it now.

What Is a Typical ‘Phoenix Race’?

Every race is different. Track conditions are different, weather is different, the cars (e.g. engine, tires, setup) are different and the competitors are certainly different in what they’ve learned since their last visit and the drivers’ emotional states.

Viewers are different, too. You see the race differently when your driver dominates than you do when he went out on lap four and the guy who crashed him is leading.

Fans and pundits like to pull out statistics to reinforce their perceptions about how ‘good’ the races were, but those statistics are often cherry picked and selected from what’s readily available. Before calling a race ‘bad’, we have to define what a ‘typical’ race is a Phoenix. To do that, I pulled the data for all 47 races at ISM/Phoenix International Raceway. I considered a range of parameters one could argue measure how competitive a race is.

Average Green-Flag Passes per Lap

Since 2005, NASCAR has reported the average number of green-flag passes per lap. This parameter provides different information than lead changes because it measures all passing, everywhere on the track. Here are the numbers from 2005 to last weekend

I’ve indicated the average over the years (5.5 passes per green flag lap) by the red line. The last Phoenix race had 4.8 passes per green flag lap, but there are some caveats to that number

• The average over all years is higher because some early races had more than eight GF passes per lap — but we haven’t see 8 since 2015.
• The average over just the last five years is 4.8 green flag passes per lap — exactly what it was last week.
• I’ve not adjusted for the fact that there were 43 cars on track before 2016 and 40 (or fewer) afterward. The average GF passes per lap goes down when there are fewer cars. For example, not as many cars are lapped with a smaller field.

Lead Lap Finishes

I was surprised last week by how few cars were on the lead lap, but it turns out that’s actually pretty common at Phoenix. Because the number of cars changed throughout the years, I compare the percentage of the starting field that finish the race on the lead lap.

• Over the last 10 years, only 45% of the field finished on the lead lap (shown by the red line). I used 10 years because the first few years were much lower than the rest of the data.
• Last week, 38% of the field finished on the lead lap.
• An additional piece of info: on average, about 85% of the cars finish the race. The lowest recent number was fall 2012, where only 65% of the cars finished. In spring 2018, 95% of the cars finished the race.

The number of cars on the lead lap last weekend was a little lower than average, but not abysmal. It’s certainly better than the fall race in 2011, when only 28% of the cars finished on the lead lap.

‘Bad’ vs ‘Average’

How do we bring all this information together to see how last weekend’s race compares to the ‘average race’ at Phoenix?

Box plots.

Some people call them ‘box and whisker‘ plots. They look simple, but convey a lot of information. Because they’re small, you can look at more variables at a time.

• The box itself shows you from the 25th percentile of data to the 75th percentile. In other words, the value of that parameter will be within the bounds of the box 50% of the time.
• The median, denoted by a red line in my plots, separates the upper half from the lower half of the data. You can think of it as a ‘typical’ value. I pinched my boxes because I thought it made it easier to see the median.
• The whiskers (the dashed lines) show the range of the lowest 25% and the highest 25% of the data.
• Outliers (which I show by green diamonds) are values so far away from the rest of the data that they can be considered exceptional.

Let’s compare the histogram and the box plot of the same average green-flag passes per lap.

• You can see here how the outliers are called out in the graph.
• The median of the data is 5.3. Half the races had more than 5.3 green flag passes per lap and half the time there were fewer.
• The middle two quartiles are bounded by 4.65 and 6.05, which means that 50% of the time, there will be between 4.65 and 6.05 green flag passes per lap.

Was Phoenix Outside the Box(plot)?

Let’s start with three quantities that might be used to define a ‘good race’: average green flag passes per lap, margin of victory and % cars finishing on the lead lap.

I’ve denoted the last Phoenix race with a pink star and put the value for that race in pink next to it.

Average Green Flag Passes per Lap (left): Last weekend’s race was on the low side of typical. See again the caveat above about the number of cars in the field not being constant.

% Cars on the Lead Lap (right): Moving to the right-most plot, this value is much closer to the median and well within the typical values for a Phoenix race.

Margin of victory (middle): We’re outside the box here, but that’s good: The margin of victory was smaller than we usually get in a Phoenix race. In fact, last weekend’s race was the 7^th closest race at Phoenix in the history of NASCAR Cup-level racing.

Three More Figures of Merit

Let’s look at three more statistics that might help us define a ‘good’ race.

Quality Leaders (middle): In addition to how many times the lead changes, the number of different drivers who lead gives you additional information. A larger number should imply a more competitive race, since more people have a chance to win.

I invented this quantity (named after NASCAR’s ‘quality passes’) to correct for lead changes during green-flag pit stops. I excluded drivers who led 1, 2 or 3 laps over the course of the whole race.

In 50% of the races, between 4 and 6 different drivers lead. There were four quality leaders last weekend in Phoenix, so that’s consistent with the lowest end of average.

Most Laps Led (right): This is the largest number of laps led by any single driver. One driver leading a large number of laps implies domination. Virtually all of the races at Phoenix ran the same distance (312 laps). Denny Hamlin, on his way to a win, led 143 laps, which is a little under the most-likely value.

Lead Changes (left): The number of lead changes (8 last weekend) is low. The lower line of the box is at 9.5. The argument about having a smaller field doesn’t hold because this stat only captures the leaders. But here’s where the histogram may help:

The average number of lead changes over all history is 13.3, but the average over the last five years is 11.3. In fact, 5 of the last 11 (45%) had eight lead changes (and one had seven) So although the number is low, it’s the same (or higher than 50% of the races in the last five years.

Conclusion

This was solidly on the lower end of a normal Phoenix race. If you were expecting ‘better-than-average’ or ‘blow-your-socks-off, you were likely disappointed and thus perceive Phoenix to have been a “bad race”.

And if you went in convinced that the new package is ruining NASCAR, you no doubt thought found reasons to support your beliefs.

Humans tend to look for evidence that supports views they already hold and ignore evidence that counters those views. We all do it, usually unconsciously. Even when you’re aware of it, you may not be able to stop it. This is why the gold standard for medical experiments are double-blind: neither the patient nor the doctors know whether the patient is part of the study or the control group, so as to minimize the possibility that either doctor or patient beliefs impact the results.

Too bad there’s no equivalent to a double-blind experiment in NASCAR.

All of this isn’t to say that there is still work to be done on the current package and NASCAR is well aware of that. NASCAR wanted to go with two different packages, but that seems to have been vetoed by the RTA because of the cost of R&D on two different packages. The change could be as simple as a different spoiler and splitter. I do not believe in the existence of a Unified Theory of Racecars that would allow one car to provide superior racing at the varied types of tracks NASCAR runs.

We’ve gotten better racing at some tracks and worse racing at others, but without viewing those races in the context of what the ‘average’ race is at each track, it’s hard to blame (or credit) the rules package. This is an ongoing experiment and the results are (as we like to say in science) only preliminary at this point in time.