I typically ride with a Garmin Edge 25 and Strava on my iphone (in jersey back pocket). The distance readings usually end up about 1.5% off, with Strava showing slightly more distance. (I'm on roads, so GPS signal generally should be good.

Any thoughts on which is more likely accurate, and why?

The Garmin is likely to have a better GPS (chipset, antenna, software stack) than a third party app polling location services in your phone's operating system.

But they'll both get your exact position wrong many times on every ride, neither one is going to be perfectly correct. Hopefully they'll be consistent. Pick one and go with it.

There's some fuzziness in civilian GPS positioning, +/- 50 meters, IIRC. Both are likely to be off.

Don't lose sleep over 1.5%. That's almost certainly within measurement error of the devices.

Warning - Engineering explanation in link
IEEE - Why Every GPS Overestimates Distance Traveled

They collect a sequence of location points and convert that to a path. They're making an educated guess as to the path, it's not just straight line segments between the points, so they might have different connect-the-dots algorithms.

Strava (or any app that collects tons of bicycle ride data from thousands of riders), in theory, could compute a more accurate guesstimation by averaging all rides from point A to point B. I wonder if they do that.

Definitely not losing sleep . . . more just curious as to which might be more accurate - there's clearly a systematic bias between the two, with Strava being consistently ~1.5% more.

Interesting article btw - thanks for the link!

You compared to the known circumference, counts of wheel rotation, of your bike computer?

My wife and I rode for 2 hours this morning. My speedo measured 21 miles while Strava said she rode 25 miles.

Measuring the distance on google maps put us at closer to 21 than 25.

I use an Android app called IPBike that reads ANT+ sensors (speed, cadence, heart rate) then allows you to upload the data to Strava. It uses the waypoint features of Strava (segments work, route map works) but uses the sensor data for speed, distance, cadence, heart rate, etc. You can also use it for directions as you ride. The user interface is a bit "clunky" but it works great. It does allow GPS only measurements, but I trust the sensors more than GPS.

If you want accurate distance and speed you'll need a wheel rotation sensor. I believe Strava uses GIS data to try and make the GPS track more accurate (there are known issues calculating distance with inaccurate GPS) so this probably accounts for some of the discrepancy, although frankly I don't think its much better. If wheel sensor data is available, Strava will use this for speed and distance I believe. Even so I usually see 0.1 mile differences between Garmin and Strava, probably a rounding error?

I encountered a discrepancy as well this morning on an early morning ride, but it was between my Edge 800 and the MapMyRide app.

With my 800 having freeze up issues, before heading out a little while ago, I downloaded the MMR app and had it running along with my 800, which thankfully didn't freeze up today (was on smoother pavement than I was on Sunday). When I got back home, the Garmin showed I did 9.13 miles and MMR showed I did 9.24 miles. Since the vast majority of my ride are less than 30 miles, what would factor out to about 1/3 of a mile over those 30 miles, if I continue to use the MMR app, I'm not going to let a little difference like that bother me.

I used a Garmin Edge 200, and now I use an iPhone 6S with ridewithgps. The phone has proven to be far more accurate for me.

I've occasionally noticed a (sometimes large) discrepancy between my garmin 800 and Strava (taking data from my Garmin). In just about every case, it comes down to the fact that the Garmin was taking revolution data while Strava was using Garmin's GPS data and my cadence/speed sensor battery was dying (or the sensor wasn't quite right with the battery anymore). Not necessarily the best way to tell that I needed to change the battery, but it worked to let me know.

I had a similar issue yesterday. Wheel sensor got bumped, and my Garmin kept jumping between ~10 and ~18mph while I was cruising on the flats.

I believe that using strava with a phone is more accurate in open roads (and road cycling). The phones these days use GPS, Glonass and data collected from cell network! On the other hand devices such as Garmin use (only) GPS & Glonass (the newer and better ones). Strava can't use a speed sensor for calculation. Cycling GPS devices (such as Garmin ones) can use data from a speed sensor (but you have to specify the exact circumference of your wheel).

As far as I know, the devices get little accurate info from cell location, just general location, nothing as accurate as GPS.

Some can be set to periodically determine the size of your wheel via gps and then use that even when the gps signal gets spotty.

Not exactly. That is the guaranteed worst case accuracy. Actual is closer to about a meter in broadcast signal. Typical smartphone accurate is about 4-5m.

See:
GPS.gov: GPS Accuracy

Lots of good information there about GPS accuracy.

Correct.

Phones use the cell tower to get an approximate position fix so that GPS positing fix is determined much faster. If you don't know where you are on the planet and the GPS has to figure that out on it's own, it takes a while. Older GPS units used to take a minute or more to get a fix. If you gave it position data that got it close, it was much faster. With cell tower triangulation, the phone has a pretty good idea where it is all the time and the GPS can rapidly get it's position accurately. Once the GPS is on and locked in position, it's going to be more accurate and there would be no need for cell tower triangulation.

J.

I have some background in geodesy and I can confirm what JohnJ80 said. The problems with using GPS, and why I advocate you use a wheel speed sensor, is due to the errors in GPS positioning and the failure to handle this errors by the software (often times because there isn't enough information, and also because it's a fairly nontrivial problem to solve). This is also why running watches are moving away from using GPS to provide distance and instead rely on a combination of accelerometers and GPS to provide much more accurate distances and pace.

Any combination of multiple measurement will increase the accuracy. that being said, The Garmin computer unit are far more precise then a app based relying on phone GPS location. I know we all want super accurate data but when you are training whether you did 22 miles or 22.2 miles doesn't really matter in the whole scheme of things.

In my ridiculously low sample size, my phone works a ton better than the Garmin I was using. The Garmin was old, and I bought it used, and the phone is an iphone 6S. What you say may be true in general, but the phone can work quite well for some of us.

I wonder when Garmin stopped updating the ephemeris for these units? Also how do you know the phone was more accurate?

[MENTION=408477]nuclear_biker[/MENTION], the Garmin occasionally showed ridiculous speeds such as 50 mph on flat ground. It also gave errors like this. I'm sure I didn't ride through buildings or out onto the river.

You don't bike in the water? That looks like a seriously bad case of multipathing. Is that NYC? Big buildings cause lots of multipathing issues in my experience.
GPS receivers have improved alot since 2011 regardless.

Depends where you ride. That's more like a best case. Rule of thumb is the nicer a place is to ride, the worse your phone will work.

You will also see seasonal variations, for example biking in the woods you will see better accuracy in the winter as the radio scattering/absorbing leaves are gone.