25

Google Maps has a bicycling routing option in beta. It does seem to do a good job routing in my town by using streets with bike lanes where possible and avoiding going up steep inclines directly.

But I'm curious how it estimates travel time. For cars, it seems to use the speed limit with a proprietary fudge factor that uses historical and current traffic conditions.

But for bicycles, what is it using to estimate travel time? There's considerable spread in terms of speed between the fastest Strava PR (personal record) seekers and the slowest bike moms with trailers - maybe even as great as 2-3x difference.

Has anyone figured out what the algorithm seems to be doing in terms of cycling times? Does it take into account the user's own cycling speed? Has Google said anything re: this?

Improve this question
8
  • Strava PR seekers? All kinds of people use Strava for tracking their cycling, and most are not PR seekers.
    – andy256
    Commented Jan 15, 2017 at 22:20
  • 4
    Yes, but PR seekers are amongst the faster cyclists on the road and bike moms amongst the slowest. Just using them as examples, apologies for the stereotyping.
    – RoboKaren
    Commented Jan 16, 2017 at 2:07
  • LOL! I read PR as something different to Personal Record. It must be Monday. Research shows more email errors on Mondays ... sigh. I'll just go away now, before I put the other foot in ...
    – andy256
    Commented Jan 16, 2017 at 5:48
  • 1
    @andy256 - I've never heard of "PR" meaning anything other than "Public Relations". I've proposed a edit to clarify for all future readers.
    – AndyT
    Commented Jan 16, 2017 at 11:24
  • I would've used PB, to mean personal best.
    – user36757
    Commented Apr 3, 2018 at 14:40

4 Answers 4

Reset to default
13

The answer given by gammapoint is surely a nice estimate to take into account when looking at Google Maps.

However, as being rather close to Google through Top Contributor and Local Guides programmes (although not being an employee), I can almost surely tell:

We'll never know.

Unless, of course, you get yourself hired in that specific department in Google.

The algorithms used by Google to estimate cycling time are based on many factors. Supposedly some of the are (the list below is nothing official, so no sources):

  • elevation
  • junctions (traffic lights or without)
  • quality and type of infrastructure
  • other users' times on similar routes

All answers on the internet will be based on estimates on specific direction queries and will try to obtain some results from comparing different or similar routes. Yet, using even bigger data to estimate behaviour of complex algorithms is difficult to the point I'll say impossible. Provided Google is using neural networks or possibly even more sophisticated algorithms, such analysis is pointless.

I'm sorry, but to find out your speed compared to Google estimates, you need to run the tests by yourself.

Improve this answer
1
  • 1
    Welcome to SE - great first answer. I look forward to your future contributions.
    – Criggie ♦
    Commented Jan 16, 2017 at 19:16
20

This site seems to have some good answers to your questions. It says

Google assumes a baseline moving speed of around 16km/hr (10miles/hr) regardless of trip distance.

but if you read more you can see there are adjustments to that baseline. For some routes where I've actually compared, I divide the Google cycling time by 1.5 to get an estimate of how long it would take me, but I used to race pretty seriously.

Improve this answer
3
  • It certainly accounts for hills, and possibly for junctions
    – Chris H
    Commented Jan 15, 2017 at 19:48
  • 1
    @ChrisH Agreed. At least in San Francisco, choosing to reverse your route can provide a vastly different estimate depending on elevation gain.
    – zahbaz
    Commented Jan 16, 2017 at 3:41
  • I came to the same conclusion by testing on my cycling. Of course if it is hilly the 16km/hr is kind of a wild assumption...
    – L.Dutch
    Commented Jan 16, 2017 at 8:11
2

Google and other traffic providers like Here (nokia) exchange traffic data, which is crowdsourced from cellphones.

If you're plotted at 5 different points along a know main road and your average speed is 50 km/h there's a decent chance its smooth-flowing vehicle traffic.

If you're riding at a snappy 30 km/h, then it could look like slow traffic. So your results get compared with other vehicles on the same road and direction at about that time. If you're roughly keeping up with the vehicles, then your trip could be detected as a car ride.

I don't know how the apple ecosystem does it, but if you use android and google services, then visit https://www.google.com/maps/timeline to explore your locations and your day's travel. Some of the travel segments could be detected as a walk, a bike ride, a train, or a car. You can correct this info too, if you see fit.

Here's the google timeline screen showing three different modes of transport, walking, cycling, and car. Also shows a wrong guess in grey (we didn't stop walking at the hospital, we just walked past it.)

enter image description here

Now all this information is available to google,

  • Average speed of cyclists on a given road
  • Your average speed over time

But google's map/routing engine doesn't use it, instead its just an average speed of the average cyclist over all roads.

I can generally travel a route in half to 2/3 of the predicted time.

Improve this answer
4
  • A good summary (+1), but note that your average speed may or may not be available, depending on your choice of hardware, settings and data connection. Similarly telling cyclists from drivers (as you mention above) or even pedestrians isn't easy on many roads (e.g. I regularly ride a couple of miles with a higher speed than the motorised average and fewer stops; there are hills in my city where the only cyclists that get up then are doing walking pace).
    – Chris H
    Commented Jan 17, 2017 at 17:19
  • ... My conclusion tends to be that Google gives a good estimate for a completely new route, allowing time for map stops and wrong turnings. With luck you'll still beat it.
    – Chris H
    Commented Jan 17, 2017 at 17:21
  • @ChrisH speed can be calculated from two timed points. No additional hardware is required.
    – Criggie ♦
    Commented Jan 17, 2017 at 19:27
  • if you know there's a straight line connecting them. And that data isn't available to Google if you don't measure it using something that connects to their servers (you could use something that has no data connection). Perhaps I should have said "...may not be available to Google etc...."
    – Chris H
    Commented Jan 17, 2017 at 22:07
0

With respect to your last paragraph/question, I get the same route and (slow) timings whether or not I'm logged in, but I have quite strict privacy settings on my Google account (in particular location history is disabled) and I don't log rides in anything that uses Google maps (my Android bike computer app - IPbike - uses openstreetmap). If your settings and use of Google apps are different to mine you may want to run a test yourself logged in vs. a fresh browser profile.

Improve this answer
2
  • 1
    Basically, you're saying that you don't allow Google to have any information that it could use to customize your timings, so it doesn't customize your timings. That seems supremely obvious and I don't see how it answers the question.
    – David Richerby
    Commented Jan 19, 2017 at 21:29
  • 1
    @DavidRicherby on the contrary they do have that information in real time. While I may not log my tracks in Google products, they know where I am while I'm riding - I'm running android with location on (and when I get somewhere their apps want me to post photos of that place, demonstrating that they know). So they could calculate my speed in real time. Turning off location history also doesn't turn off very much and may not be applicable given how much latitude they give themselves.
    – Chris H
    Commented Jan 19, 2017 at 21:48

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