Google Maps: erratic "In Current Traffic"

Where'd it go?

It shows on my Android phone.

The Shape of My Commute, part 5

A simpler chart. This is only directly observed estimates (red line) vs. actual times (blue line). The point at noon is "fake" so that the shape of the commute is easier to see.

I still feel like the accumulated, averaged estimates from Google maps are too low, but that may just be the effect of the bad days having more psychological impact. This chart below shows directly observed estimates (red line) vs. accumulated, averaged estimates (blue line). I presume that, after some number of direct observations, the red line could be smoothed to something that would closely match the blue line, but I don't know enough about statistics to figure out what that number of observations would be.

Or maybe the commute is really getting worse.

Here, also, is a link to the Google spreadsheet with the data.

The Shape of My Commute, part 4

More actual data. More questions.

The Average Estimate numbers from Google Maps for the morning look slightly optimistic (5% - 10% low).

The evening numbers are much farther off, more like 25% low.

Don't know if I'll ever know why. I can, however, come up with a "realism factor" that I can apply to the Google Maps numbers to get something closer to my Actual observed numbers.

The Shape of My Commute, part 3

Starting collecting actual data. After three data points it's not looking good for the estimated times, but confirming my gut feel that the estimated times are too low. I added an "actuals" column of data (shown in red). The spikes show the data points I've collected.

Still, from the estimates I can get the overall shape (e.g., commute times at each slot relative to each other), even if the absolute values are wrong. Overlaying the actual data will then give me the picture I need.

I'm guessing that we will see a wider variation between estimated and actual close to the peak times. We'll see if that guess proves out as I collect more actuals.

What's The Next Big Thing?

www.google.com

http://blogs.denverpost.com/techknowbytes/2013/03/12/whats-the-next-big-thing-in-technology-lets-ask-sxsw-interactive-2013-attendees/9076/

 www.yahoo.com

 http://graphicsweb.wsj.com/documents/NEXT_BIG_THING/NEXT_BIG_THING.html

 www.bing.com

 http://www.pbs.org/wgbh/nova/tech/what-is-the-next-big-thing.html

www.wolframalpha.com

http://www.wolframalpha.com/input/?i=Next+Big+Thing+the+album&lk=1&a=ClashPrefs_*MusicAlbum.NextBigThing%3A%3A1ec1d438!-4621!-3769!-8ab4!-2d5590fbd44a-

The Shape of My Commute, part 2

After collecting data for a couple of weeks, and throwing out the worst of the bad data days (why does Selenium in Firefox crash out so often? to be determined...), I now have this:

The blue line is my commute tolerance (currently at 40 minutes).

The actual numbers (e.g., Google Maps estimates) feel low. I realize that time spent sitting in traffic seems longer than it actually is, but I would have guessed +5 minutes at peak times.

In the past I've tried to keep track of my actual commute times, but lost patience after taking a few samples. Maybe it's now time to revive that practice and do some "reality vs. Google maps estimates" validation.


Part 1

The Shape of My Commute

I've been commuting from the east bay to the peninsula for many years, to many different locations. I realized that I've never had any reliable data about the best and worst times to leave home or work.

I decided to use the estimated times that Google Maps provides to try to get a clearer picture. I first wrote code to use the Google Distance Matrix API to get estimated driving times. I got that working, but got a nice (but useless) flat line when graphing estimated commute duration by departure time. I now assume that the API provides the first number shown on the web page, not the second:

I then re-wrote my code to use Selenium to grab the data off the web page. Those numbers made more sense. I don't have many days of data yet. I'm assuming that, as I get more data and average it in, the bumps in my graph will smooth out a bit.

If you want to do this, my code is available from Github. You will need to download a few Java jar files to get it to run.