Archive for November, 2013

Time to look at some gripping data, statistics, and data presentation methods, along with a rare chance to use the word ‘oxygenated.’

I drive a 5-speed Honda Civic EX which I bought new in 2006. Not too long ago I looked down at the dashboard and saw the big 100k miles approaching. Somehow the following picture managed to show up in my cell phone, obviously taken by a professional driver on a closed off road. And not by a simple-minded commuter who happened to be in the middle of heavy freeway city traffic and going the indicated 60 mph at the time.

Don't try this at home. Because it's hard to get a car going 60 mph inside.

The odometer shows 99,999 miles, so I missed the 100,000 mile milestone by one, but there was only so much driving-by-knee in heavy traffic that I could do that day.


But this post is not about odometer / speedometer pictures. It’s about something even more fascinating: gas mileage. (“Wow!”) I’ve been methodically collecting mileages and gas numbers at each fill-up over the last 7 years. (“Why??”) I managed to gather up all the numbers over that time, calculating and compiling mileage numbers tankful by tankful. Here’s a graph of the car mileage over that time. Click/tap on the picture for a larger (more fun!) version.

Miles per gallon (MPG) for each of 298 tankfuls over 7 years.
Vertical color bands are by season of the year.

A few comments on the graph:

  • The blue spiky line (with diamonds) is the mileage for each tank, varying between 28 and 40 mpg.
  • The orange line is the average mileage over time (total miles divided by total gallons). It slowly increased and settled to about 33.7 mpg.
  • The wavy green line is the average mileage over the last 10 tankfuls at each fill-up. That is, it represents the mileage of the most recent tank at a given fill-up date, and the 9 fill-ups before that.
  • The vertical color bars (green, yellow, orange, purple) show the season of the year (spring, summer, fall, winter)

One interesting part of the graph is the green line with the triangles, which gives a 10-tank running average of mileage. The tank was filled on a roughly weekly basis, so each point reveals the mileage over the previous 2 – 3 months. This line always peaks in late summer to early fall, and its recurring minimums occur each winter. I get around 8% to 10% difference between the high and low peaks.

Some other questions / insights one might have regarding the mileage graph:

  • Why does the overall mileage (orange line) slowly but consistently improve from about 32 mpg to 33.7? Does the car ‘break-in’ or get otherwise happier over time?
  • The worst of single tank mileages (blue points) almost all occur toward the winter, with an occasional bad one in the spring/summer.
  • The top half-dozen tankfuls all occur in summer, but there was one good one (>37 mpg) in January of 2013.
  • What causes the seasonal cycle — higher mileage in summer, lower in winter?

Weather Impact?

For that last point, on the seasonal variation: One cause of course might be temperature change having a direct impact. Obviously that follows the same pattern of high-in-summer, low-in-winter. There’s lots of data available for the curious on various historical weather effects, for example from this Weatherspark site.* The graph below is temperature data for my area over the same time period, 2006-2013.

Global warming / cooling is staying carefully concealed over this period.

Temperature data for Seattle (Boeing Field) for 2006 – 2013.

Incidentally and as a slight aside, this site also allows exploration of cloud cover statistics. Here’s how things looked (in the sky) for the same time period, 2006-2013 (that’s 0 to 100% on the vertical direction, for the shaded gray graph):

But it only rains at night

Cloud cover stats for Seattle, 2006 – 2013

So apparently there have been a few clear days here and there in the last 7 years. And not to depress the locals or anything, but here’s how the cloud cover numbers (or lack thereof) look for Los Angeles over the same time period:

Yes but it's only sunny during the day

Cloud cover stats for Los Angeles

Anyway it doesn’t look like cloud cover statistics is going to help account for the seasonal variation in gas mileage. Some other reasons are proposed at THIS SITE, and include the following:

  • Colder fuel doesn’t explode as easily (because it doesn’t atomize so well)
  • Oxygenated fuels, often applied in the winter months to improve air quality, reduce mileage by a few percent
  • Changes in driving style, which can be an indirect effect of the winter climate and planet orbit. More darkness and wet roads leads to slower speeds and more congestion, which reduce mileage. My commutes are frequently longer in winter for those reasons.

Data Presentation

Let’s use this data as an opportunity to explore how to bias or distort data to communicate a point you want to make. One good trick is the careful use of Y-axis manipulation, i.e. scaling the vertical axis to suit your purposes. Among the 298 fill-ups, I broke out the top 8 station brands** that I used (for 279 of the purchases), and then shifted the data so the mileage from each station could be compiled correctly (using the excel pivot table function).

This first chart shows the average mileage from each of the 8 stations. Note that the vertical axis starts at 0. The normal passerby would probably say that all of the stations give about the same mileage — it doesn’t really appear to matter much where you buy your gas.

All gas is created equal, but some is more equal than others

Mileage data by gas station brand. All the same, right?

But if you want to bias the audience toward or away from some station(s), how about changing the vertical axis (without changing any of the data). This time (next graph) just show the mileage numbers starting at 32.80 mpg. This is the exact same data as the previous chart, but now a quick glance from that average passerby will get a different reaction. It looks clear that Safeway gas is for losers, while Union 76 is the champion — why, maybe it’s 20% better than the others! But in fact, the Union 76 mileage in my data is less than a half-percent better than Shell, and only 4% better than Safeway.

But zoom in far enough on small differences, and they can look big.

Zoom in on average mileage per station, with vertical axis starting at 32.8 mpg.
All wildly different, right?

Another thing not revealed in these plots is how many purchases were made from each place. A small data set for something like mileage tends to vary more than a large collection, which would be more settled toward its average value (if there is one). The next chart changes the vertical axis again, to start at 32.0 this time, and also adds some color and thickness effects to try to show frequency of visits. But on the very likely chance that this doesn’t work, the labels on the bars makes it kind of obvious.

Similar graph with vertical axis now starts at 32 mpg. Also adds number of time each station brand was used.

Yet another vertical axis starting point. With ‘depth’ and color added to the columns in a lame attempt to indicate frequency of purchases.

This still can be misleading because, for example, perhaps all the Safeway fill-ups occurred during the winter when mileage is typically lower as discussed above. But a careful look through the numbers shows that 7 of the 9 Safeway visits occurred during the spring and summer. Which means that Safeway is actually biased somewhat more favorably than deserved in these graphs, and it still loses. (By a little. Or was it a lot?)

Gas Prices

Finally let’s see how gas prices have varied during the last 7 years. This next graph might lead one to the following creative conclusion: gas prices above $4.40 are a strong indicator of imminent economic collapse. Maybe there’s something to that. If you’re so inclined, you can look up gas prices at your location (US and Canada) during this century HERE.

Gas price paid over the years

Gas price paid over the years

Wrap-up on Fill-ups

Some summary numbers for my 298 fill-ups over this period:

  • Total gallons: 3119
  • Total gas cost: $9863
  • Average cost per gallon: $3.16
  • Total miles to this point: 105k

* That site allows for interactive modes, click on its ‘View in dashboard’ option
** FMeyer is Fred Meyer, a local-ish department store sometimes including gas stations.


Read Full Post »

Netflix on PS3

Netflix on PS3

While I work in technology fields and am surrounded by and use devices all day long, I was recently stumped by a seemingly simple problem: how to change users in Netflix? A relative had come over and changed our login to his account, and I wanted to change it back. We get Netflix to our TV through a PS3 system and its wireless connection to the internet.

This turned out to be a surprisingly hard problem. I may be slow, but I couldn’t find anywhere on the Netflix user interface for logging out.

Solution? My friend Google led to this very short video. The video runs a whopping 36 seconds, but the useful content only takes about half that time. And for some strange reason, the audio tends to play faster than the video on some systems and browsers: I checked on PC and Mac, Firefox and Chrome, and it had the audio running fast. But on my tablet it played normally. (So the useful audio may be over in 12 seconds if that’s playing fast.)

Anyway, the technique for getting to the logout screen, shown in the video, was odd enough that I thought the video might be a joke. But I tried it and it really works. Just scroll through the movie icons in this pattern: up – up – down – down – left – right – left – right – up – up – up – up. Would a sane person expect this to work? (This probably became public as part of the wikileaks revelations, and beware that a bell probably rings at NSA headquarters whenever anyone successfully completes the pattern.)

Minor usage note: On a PS3, this is more reliable if you use the ‘arrow’ keys rather than the joystick — the joystick can work but doesn’t always, perhaps because it senses some slight ‘angles’ which the interface uses as a straight scroll, but the pattern-recognizer rejects those to protect the super secret logout access.

Go ahead and invest / waste 36 seconds of your life on the video, and another minute going off and discovering that it actually works. Here’s how it looks for those who like numbered instructions. For some reason.

As simple as 1-2-3-4-5-6-7-8-9-10-11-12

The secret pattern revealed. After this, the account page magically appears. Really.

Read Full Post »

%d bloggers like this: