14

I have this route planned in Google Maps as a rough plot for travel times to 4 locations from Tokyo Tower, it's by no means finished.

Now I am well aware that there are other factors such was road works, traffic conditions, rest stops, etc. but would the travel times listed on Google Maps be an accurate estimate?

i.e. From Tokyo Tower to Hachiman Shrine, can I expect it to take at least 14 h 28 min on a perfect non-interrupted drive with the route Google Maps suggests?

  • 1
    I stopt using Google for predictions. Crowdsourced tools that are based on a decent crowd are better sources. Currently I am quite enthousiastic about Waze for traffic predictions – user141 Dec 7 '14 at 9:00
  • 3
    Psst: Waze is owned by Google :) – lambshaanxy Dec 7 '14 at 23:12
18

I've found (anecdotally) that initially in London, the walking times were way too slow - I was beating the times regularly.

Then I moved to Vancouver and found them too fast. I'm a quick walker, so wasn't sure what was happening.

I eventually figured it was down to knowledge. I 'knew' London far better, and even though I might be using a map, I could navigate the crowds, the streets and so on far better than in Vancouver, where I was regularly having to check landmarks.

When it comes to driving, the Google Maps times are best possible times. So it's assuming you can drive at basically the speed limit all the time. It gives no thought to inevitable traffic problems (they're improving that, especially in cities, but it's not there yet), and as you've pointed out won't handle rest times.

I drove from Melbourne to Sydney (Australia) this year and the times were definitely low. I wanted to stop for lunch, or briefly check out some sightseeing spots, sure, but even the sections where I was just driving (week day, during the day) in a car I wasn't used to, on roads I wasn't used to, the estimates weren't ideal.

Perhaps if you speed, or are really used to driving on Japanese roads it'd be more accurate, but currently it's a best-case, for the most part.

(disclaimer: this post will outdate as Google improves, but that's the situation to the best of my knowledge currently)

EDIT: Someone did a pretty good breakdown, and a discussion follows, on this topic on reddit (it was about why speeding seemed to match the accuracy). Some of the discussion will be of interest:

ELI5: When I get driving directions from Google Maps, the estimated time is usually fairly accurate. However, I tend to drive MUCH faster than the speed limit. Does Google Maps just assume that everyone speeds? How do they make their time estimates?

Further relevant posts, linked from that discussion:

Speed Limits: How does Google maps calculate your ETA?

and how they're trying to use your data to improve their algorithms:

Google Maps adds traffic data from your cell phone

  • 3
    "it's assuming you can drive at basically the speed limit all the time". From my experience it seems like it's assuming you're driving slightly over the speed limit. – vartec Dec 4 '14 at 4:41
  • 1
    Google maps predicts traffic. When I came home from work today it said something along the lines of 16 minutes to home (2 minute traffic delay) (I don't remember the specific wording, sorry). It also recognises accidents. – Jon Dec 4 '14 at 6:13
  • 1
    I believe that a lot of Google Maps' live info comes from data from users with GPS (I could be wrong, just something I seem to remember a Google Maps designer mentioning during a presentation). So, London estimates may be slower than Vancouver because London's average speed is dragged down by slow-moving tourists, and maybe Sydney commuter traffic is under-estimated because it's mostly solo drivers on a known route, not using GPS and therefore invisible to Google's data centres? – user56reinstatemonica8 Dec 4 '14 at 13:00
  • 1
    @IvoCoumans if that were the case, the times would remain constant. As seen in the links provided, this is not the case, and far more than just basic maths is involved. – Mark Mayo Supports Monica Dec 4 '14 at 13:45
  • 1
    @IvoCoumans Here's a data point with regards to GPS. I was recently driving between states, and I noticed that Google Maps (not navigation) on my Android device claimed it was going to take 4 hours for what should have been a 2 1/2 hour drive. Due to cognitive dissonance I dismissed this, but about 10 miles outside of town I ran into a standstill in traffic and it took nearly 30 minutes to get a mile down the road to an exit. So at least in some situations, Google does take recent GPS data into account. – user23030 Dec 4 '14 at 21:05
13

Good, but not perfect. Any planning tool can only give you estimates based on past experience, but predicting the future is by nature fraught with risk. For reasons explained below in detail, Google Maps is currently less accurate for this route in particular and Japan in general than it would be for, say, the SF Bay Area in the US.

As Mark points out, estimated times tend to be "best estimate", which assume you can drive at the legal limit at all times and never take rest breaks. In Japan, driving at the limits is actually more realistic than in many other places, since Japanese speed limits are generally set absurdly low (often 80 km/h even on expressways) and consequently more or less everybody speeds.

On the flip side, at time of writing driving directions in Japan do not incorporate current traffic information, which can be misleading particularly in big cities like Tokyo. Even some of the major expressways, eg. the Tōmei between Tokyo and Nagoya, are notorious for traffic jams at peak times/seasons. (Update: Traffic is now accounted for.)

You've also got a whole bunch of ferries in there, whose schedules are not incorporated into the planning, and the big ones (eg. Oma-Hakodate between Honshu and Hokkaido) only run twice a day or so.

I've got to ask, though, are you sure you want to drive this thing from end to end? Tokyo is a nightmare to get around by car, and many of the expressway stretches (eg. Tokyo-Osaka) are both deathly dull and expensive due to heavy tolls. Using Shinkansen bullet trains when possible and only renting a car at your destination if needed is likely to be faster, more comfortable and cheaper if you make use of the JR Pass.

Disclaimer: I work on Maps at Google, although not on driving directions specifically. This answer represents my personal views and not those of Google.

Update since my original lead of "As good as it can be" seems to have sprouted a debate in the comments: all I meant was that Google Maps does a good job of planning, given its inputs. But of course it can't account for your walking speed, or model of car, or the driver's spastic colon that necessitates changing a colostomy bag every 15 minutes, because it's not told about those. Should it? Maybe yes, maybe no. In the aggregate, though, I find it good enough already, and am not aware of meaningfully superior alternatives for general applications.

  • 1
    I think "as any planning tool can be" is overly optimistic and generalising. A planning tool could ask you for how fast of a walker you are, or it could look at the average walking speed of previous tracks, or it could present times for both a slow and a fast walker, or or or. I found that "it's not improvable" is never true, and often even counter productive. – phresnel Dec 4 '14 at 9:21
  • @phresnel Whether option creep is an improvement is debatable. For a general public UI, it's generally thought to be a bad idea. – Relaxed Dec 4 '14 at 12:56
  • @Relaxed: Those were just examples. A better tool may also take into account different travel times at different times of day. E.g., in my hometown, traffic lights have different day and night modes; thursdays are faster than mondays; got two rush hours; saturday mornings there's a lot of old folk driving way below the limit, etc. etc. These are just some more examples to counter your "generally thought" and "option creep" "arguments", which are pretty lousy arguments against my core doubt (which is that "as good as possible" is always an untrue argument, until general perfection is reached). – phresnel Dec 4 '14 at 15:30
  • @phresnel It's not a lousy argument but a major design consideration… What's counterproductive is getting hung up on small details like the ones you mentioned. My point is precisely that “perfection” is a distraction, good design is about trade-offs and taking into account quantitatively negligible details at the cost of greater complexity is not an improvement. – Relaxed Dec 4 '14 at 15:36
  • @Relaxed: A major design consideration is to realize that it's the small things that accumulate to the overall result. If you drive through Düsseldorf on a Tuesday night, you can make all your way through within maybe 15 minutes, from west to east. On a saturday morning (in Düsseldorf, it's shopping folk instead of old folk), you need to be lucky to make it in an hour. Or when travelling from Düsseldorf to Hamburg, a number of small construction sites can increase your travel time from 4 hours to 8 hours. If the target of your software is reliable travel planning, I utterly fail to see why ... – phresnel Dec 4 '14 at 15:41
3

I've found the estimates pretty accurate in the USA, to the extent that, when I've taken photographs from the passenger seat on road trips, I can often find out exactly where a photograph was taken based on "Seven minutes ago, I took a photo that randomly included mile marker 123; Google maps thinks seven minutes from there takes you to here; Street View matches perfectly." Obviously, that's a pretty short distance but longer journeys seem quite accurate, too.

3

I cannot answer for Japan, but in Kuwait Google is by no means exact, but it has been fairly close (~5 minutes from my anecdotal calculations) to the actual time it took between destinations.

In Kuwait it also accounts for traffic congestion and will tell you how much delay traffic is causing on your commute. It uses a combination of data points. I know that it uses including cellphone tower data as there was a delay in getting approvals for this and for a while we had maps, routing but no "traffic" information. I am sure it uses multiple data points in aggregate to create an estimate.

In Dubai it has transit information as well, and it has also been very accurate (especially in the wait times between the train stops and the time it takes to walk between stops).

However, it is not the fastest way to get between two points, because it lacks (at least in Kuwait) the ability to re-route based on traffic. So if you know the roads you can always find a faster way. I suspect in Japan it may not know all the best ways to get from A to B.

  • Not quite sure what you mean by "reroute based on traffic"? Maps will offer multiple alternatives if the obvious route is congested, and Navigation (mobile only) does reroute based on traffic. support.google.com/gmm/answer/3273406?hl=en – lambshaanxy Dec 7 '14 at 23:12
  • I know, but this automatic rerouting is not available yet in Kuwait. – Burhan Khalid Dec 8 '14 at 2:11

Not the answer you're looking for? Browse other questions tagged or ask your own question.