How far can we go after the gas light turns on?

Ever wondered how long you could actually drive after your gas light turns on? Well, Justin Davis at http://www.tankonempty.com might have an answer for you. Justin has been collecting reports from daredevil drivers about how long they’ve driven post-gas light, and he was kind enough to give me his data.

Test
“Listen to me! When that car rolls into the dealership and that tank is bone dry, I want you to be there with me, when everyone says: ‘Kramer and that other guy, oooh they went farther to the left of the slash than anyone ever dreamed!'”

The dataset is pretty extensive, with records for every make and model imaginable—some that I’ve never even heard of! Ever seen a Vauxhall Vectra? It looks pretty unremarkable actually. They go about 32 miles after their light turns on.

Most of the data comes from cars made between 2005-2010, but records stretch back as far as the 70s.
Most of the data comes from cars made between 2005-2010, but records stretch back as far as the 70s. (Actually there was some data from older cars, but too few entries resulted in some unreliable estimates.)

So how long can you drive after the gas light comes on? The values provided by the drivers are remarkably consistent actually, with the mean value hovering between 30-40 miles. So TL;DR? You can go about 35 miles.

This is a hexy heatmap; the blue is a smoothed best-fit line with a shaded region for standard error.
This is a hexy heatmap: the redder an area is, the more values are contained within it. The blue is a smoothed best-fit line with a shaded region for standard error (it gets wider as you go back in time because there are fewer entries for those old cars).

But if you look closely, you might notice something fishy. Starting around 1995 we have a bunch of drivers reporting the max value permitted by the website: 99 miles! Let’s look at the distribution of values…

Remember, these are all self-reported miles driven. A fitted normal curve helps us see where we have more values than we might expect.
Remember, these are all self-reported miles driven. A fitted normal curve helps us see where we have more values than we might expect.

We get two things from this: first, we see that people tend to provide ’round’ values (like 30, 40, 50) rather than values in between (see: avoiding false precision). Second, we see that there are way more 99 values than we might expect. This either means that the website isn’t built to accept a large enough range of values, or—and I think this is more likely—that most of the 99 mile values (if not all of them) are bologna. So who’s submitting these 99s? Honda, Toyota, and Ford owners account for almost half of these fishy values, but when you control for base rate, Hummer emerges as the biggest fibber. In fact, over 1/3 of the entries for Hummers are at 99 miles!

“Do you think maybe he’s compensating for something?”

Maybe you want to see how your car stacks up. Because we have a set of estimates for each car, we can plot them according to 1) the average distance people report going, and 2) how variable those reports are. This gives us a cool plot with four distinct regions: short and unreliable, short and reliable, long and unreliable, and long and reliable.

The red and blue lines represent the average values for the standard error and mean distance, respectively. The text had to be pretty small, but... can you find your car?
Cars in the top right are what you might call the “unreliable boasters,” while cars in the bottom right are the “old faithfuls.” The text had to be pretty cluttered, but… can you find your car?

Now lastly, the entry form also recorded the users’ IP addresses. Thanks to some simple geocoding website I found, I converted these IP addresses to latitude and longitude. So just out of curiosity, let’s see which states these data are coming from!

Mostly from Louisiana. Wait! There are only a few cases here.
Mostly from Louisiana. Wait! There are only a few cases here. So where are the rest (10k+) coming from?

From a sample of over 11k, the densest point in the US has four respondents. That’s odd—who’s giving us all the data then? Zooming out to the whole world lets us find out.

Looks like there are a lot of Turks that like reporting their gas mileage.
Zooming out to the world map, we see a more complete picture.

The Turks! Yep, it seems that the vast majority of respondents are in Turkey.

“If you go to the grocery store and stand in front of the lunch meat section for too long, you start to get pissed off at turkeys. You see turkey ham, turkey pastrami, turkey bologna — somebody needs to tell the turkeys, ‘Man, just be yourself.'” -Mitch Hedberg

It would seem that the Turks (particularly the ones in Ankara) love the tankonempty website. So do the findings generalize? You be the judge!

Want more data visualization? Check out my other posts at: https://vizthis.wordpress.com/

Advertisements

3 thoughts on “How far can we go after the gas light turns on?

  1. I’ve often thought about that, especially since various cars will find different ways to let you know how much gas you have left – some of them even ‘lie’, so to speak. They even turn on the light early on, or way too late – it’s a fun experience; if you get to change cars often, you should try it out. It’s fun when done on purpose.

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s