The Strava Struggle; GPS Accuracy

For those of you who don’t know, Strava is an online “social media” program designed to let you keep track of your runs and bike rides. Not only does it maintain a log of your mileage, time, and intensity, it also allows users to create “segments” on their routes and turn any piece of geography into a virtual time trial open to all comers. Strava takes a user’s bike ride, recorded through any number of GPS devices, and sifts through the data to see if the cyclist crossed any of the invisible start/stop lines demarcating a segment. This is used to generate a “lap time” of sorts which is automatically compared to every other cyclist that rode the same segment and stacks up the times into a virtual  leaderboard in which the current record holder receives the coveted status of King Of the Mountain. (KOM) This sort of online contest is especially appealing to road cyclists who are just as prone to narcissistic competitiveness as they are to shaving their legs and wearing Lycra. I am certainly no exception to this phenomena, and in fact, years of amateur level motor-racing primed me to fixate on leaderboards like a hypnotized lemur.

Like a lot of people, when I first started using Strava as a data-logging/online dick-waving match, I was recording my rides with a mobile application run through a smartphone. Strava found the mobile app was a great way to market their product to a large audience, and a lot is to be said for how streamlined and easy this makes it to upload rides while you flagrantly violate The Rules by listening to U2’s latest shitty album release through headphones. Unfortunately, early on I began to suspect there was something to be desired in the GPS accuracy attained via mobile phone. It seemed the data I was uploading produced a virtual map of my route that looked a lot more like a QRS segment on an EKG machine than a guy riding a bicycle down the road.

Is it a normal sinus rhythm or me riding the Huckleberry Trail?

Is it a normal sinus rhythm or my curiously creative route to the local Starbucks?

 

This was no exaggeration, data rendered from my admittedly old Iphone 4S produced a “route” which showed me impossibly zig-zagging across rivers, riding through trees, or phasing straight through buildings. None of these abilities were ever advertised as being stock features on the Specialized Allez Sport I’m riding, so don’t break your spindly little cyclist ankles rushing out to buy one.

Apple catching me red-handed using alien technology to create my own shortcuts.

Apple catching me red-handed using alien technology to create my own shortcuts.

Naturally this made me a wee bit concerned, but naively I assumed Strava must’ve been using some sort of genius computer algorithm far beyond my feeble understanding to sort out the relevant data points and tease pretty accurate segment times out of the mix. Sure on cloudy days the damn thing might be missing every single segment along a particular route, but when it found one, it must’ve been roughly on par. I rationalized all GPS systems are crude to a degree, and since Strava only measures to the nearest second, there’s a certain degree of accuracy lacking in the first place. I knew there had to be some errors taking place on specific segments, but surely this was just confined to short routes through underground tunnels in radio jamming zones where the GPS coverage was even spottier than usual, right?

Wrong. One day while cruising a route at a brisk pace that was far removed from a full blown hammer-fest at maximum heart rate, Strava informed I just nicked 2nd overall on an uphill segment at 22 MPH average speed. This couldn’t be right, my cheapo replacement for my beloved Cateye might be a shitty speedometer, but I was quite confident I had never once touched 22 MPH even on the flat leading up to the climb. Luckily for me, my dad supplied me with a shiny new Garmin 200 Edge a couple days later, and I now had the perfect opportunity to compare the data logging between my now highly suspect Iphone and a dedicated GPS device. Just toss the ol’ cellphone in the jersey pocket and slap that Garmin on the handlebars and away we go.

My first ride was both troubling and reassuring, as I smashed my old KOM on the local steep hill I’ve hated since I was 10 years old and snagged 3rd place on a fairly challenging set of rolling hills. This data was recorded through the Garmin, and it showed a neat, clean little map of exactly where I had ridden with zero deviation and vindicated that I was indeed capable of producing a few watts without aid from glitchy Apple devices. Meanwhile the Iphone was missing entire segments, and displaying a spaghetti mess of a route with segment time discrepancies of over 5 seconds. Uh-oh. My dad replicated a similar test with his Android, and while the Droid faithfully rendered a close overlay of his route and even got the elevation right, it also estimated he was 1-2 seconds faster on segments than the Garmin. In racing terms, beating someone by 2 seconds is colorfully referred to as brutally spanking the competition like a helpless puppy, but for amateur cyclists on small leaderboards it can be irrelevant with times separated by over 10 seconds anyhow. (While I definitely remember competing in auto-races where the difference between 2nd and 3rd came down to thousandths of a second, we’re just dorky bicyclists without the benefit of transponders and timing boxes)

Still, as I began to look back at the local leaderboards on the most popular local routes, I noted they were curiously dominated by Iphone users. Moreover, there were huge gaps between the KOMs and the next closet rider, far in excess of a few seconds. To confirm my sneaking suspicion that smartphone data was producing downright hilariously bad and outright fantastical speed estimates, I picked a very popular local trail with multiple segments to cross compare with. What I found pretty much destroyed the credibility of bike rides recorded on an Iphone:

Segment times rendered from Iphone data.

Segment times rendered from Iphone data.

 

Garmin data from the same exact ride.

Garmin data from the same exact ride.

While some of these segments are close, or even identical, two of them are estimated as significantly faster on the Iphone, one by 9 seconds, and another by 24 seconds.  Now that’s just plain ridiculous, and it makes any KOMs, Personal Records, or Hell, just average speed and distance in general completely unreliable. (My Iphone regularly over-estimates average distance by over a mile, presumably from all the Crouching Tiger, Hidden Dragon ninja moves I’m doing in the tree tops) Getting variances of half a minute can mean the difference between dragging a digital boat anchor or suddenly transforming into Eddy Merckx.

So what’s the take away from all this? First, most obviously for us competitive, humorless assholes, it means any KOM from an Iphone can be immediately sneered at as pure cheating, probably intentional, and definitely by some Fred bike-pathlete that can’t even ascend the local Cat 3 climb without stopping to take a nitroglycerin tablet and inhale an asthma medication. (In all seriousness, most people probably don’t realize how bad the data logging is) Secondly, it means using an Iphone for training is probably an exercise in futility since it’s going to be randomly adding and subtracting 100 watts from your performance on various segments. If you’re going to seriously ride a bicycle, you need to ditch that smartphone and those ridiculous pedal straps. Finally, it certainly seems to suggest Strava needs to seriously do something about the leaderboards and the algorithm being used for smart-phone data, especially those smart-phones with a popular fruit stamped on them.

Supposedly the problem lies in the sampling rate of the GPS systems in cell-phones, particularly older models that lazily connect to the satellite every half hour or so. Strava tries to just line up the data points which are closet to the segment, which can conveniently produce a shorter, faster route over the given stretch of distance, or a frustratingly longer and less direct path depending on your luck at Iphone Roulette. Given that so many highly optimistic KOMs are poised to be lost if Strava alters their algorithm and applies retroactive changes, I can see why there would be some hesitancy about making thousands of users lose their top dog status overnight. Also, equally annoying is the fact segments themselves are often based on glitchy zigzag patterns, and that users driving or riding on different routes nearby can be matched to the wrong segments. (Best example of this was someone on a highway bypass popping up on the leaderboard of a local bike path)

Until the technology and software catches up, I’d stick to dedicated GPS units only.

Advertisements
Posted in Uncategorized | Leave a comment

Hello world!

Welcome to WordPress.com! This is your very first post. Click the Edit link to modify or delete it, or start a new post. If you like, use this post to tell readers why you started this blog and what you plan to do with it.

Happy blogging!

Posted in Uncategorized | 1 Comment