Why you will never run a marathon on a treadmill
There was an interesting article in the press recently leading up to the London Marathon. The Which? consumer magazine tested over 100 fitness trackers on a calibrated treadmill and announced huge variances in performance - naming brands that incorrectly reported the marathon distance by over 10 miles. But how relevant is the test they performed?
Which? state that they tested the brands over a 2 year period using a calibrated treadmill. But is running on treadmill with a fitness tracker a good proxy for running a marathon with a fitness tracker?
No.
The most advanced, and highest-performing fitness trackers will be performing sensor fusion, combining a GNSS receiver with measurements from accelerometers, gyroscopes, magnetometers and barometers. The device will be programmed to assume that a runner's stride length, gait, and running style will vary between different people. With their level of activity and exertion, the tracker will adjust its model of human motion accordingly. It will also bypass the GNSS receiver whenever the satellite signals are blocked or corrupted, and bridge the gaps using just its tailored human motion model. A treadmill forces the runner to maintain a fixed pace, something that is not possible in reality when faced with variations in terrains, turning corners, and moving around other runners and obstacles. All in all, there are a great deal of differences between pounding along on a treadmill indoors, and running freely a great distance outdoors.
Critically, fitness trackers cannot do any adaptive modelling while running on a treadmill, without actually moving through space under the coverage of GPS signals providing an external measurement of distance travelled. It is therefore unfair to perform the tests in this manner when devices are not designed to be indoor treadmill trackers.
So we really need to look at test data from actual outdoor running - and conveniently data from marathons of exactly this type is actually available. This study showed that GPS based trackers typically exhibit an error range of around a mile across the full distance with the average distance being recorded being around a quarter of a mile longer than the true distance. This is unsurprising as GPS traces all contain some degree of "wander" around the true true, and the total distance of any wandering path will typically longer than the intended less-meandering path.
So the Which? article - or at least the press resulting from it - was a little unfair, there is rich evidence to suggest that it is very unlikely that over a full marathon distance out in the open fitness trackers really do result in errors of 50%. But errors of 5% are still unacceptable.
And so which one really is the best fitness tracker? Well they all contain the same components - GPS receiver, accelerometers, gyroscope, etc. Their performance in reality comes down to some quite basic questions - how sensitive is their GPS receiver, and how good are their sensor fusion algorithms?
All fitness trackers can be greatly improved with Focal Point's software - we can improve the sensitivity and accuracy of GPS receivers by over a factor of ten, and have a tailored sensor fusion system specifically for human motion tracking designed by world experts in positioning and navigation. Focal Point are working at the moment with major players in the industry to deploy our software in consumer devices in 2019, so watch this space...
If you'd like to learn more about D-Tail or S-GNSS, take a look at our tech page or drop us an email on info@focalpointpositioning.com