I don't think Footpods are more accurate on the road. They may be, but I only need one for running on a treadmill or indoor track.

The protocol behind the study in that link is defective for assessing GPS device accuracy. And, it is optimized to favor foot pods.

The better foot pods are using 3-axis or 6-axis accelerometers to map movement in 3-D space and then translate that movement to distance and velocity. As noted in the above protocol, Stryd is doing that very well.

The TL;DR summary of accuracy is that footpods are accurate in shorter intervals, because they have a small relative error with each step you take. GPS is better at longer intervals, because GPS has absolute error that tends to cancel out over distance.

Then there is also the issue of data smoothing; which different devices seem to handle differently with GPS signal (not sure about the foot pods), in addition to time that it takes to acquire a GPS signal (you just want to start running, and lose a couple miles of run data because you did not stand still to let it 'acquire' gps signal), whereas the foot pod is ready to go when you are (as long as you turn it on and sync with your watch/device). Add to that the indoor training on a treadmill where your GPS signal will not move very much...so the foot pods don't really beat GPS, they are just needed to augment it.

---

I believe my local reality has been violated.
____________________________________________
Happiness = Results / (Expectations)^2

Ha, you are nailing the Apple Watch here. :) My favorite quote about the AW "it just counts to three like a toddler and hopes for the best."

Short answer: They fudge it based on processed acceleration data. They can be very accurate across a range of speeds but not for everyone.

Long Answer: Foot pods have accelerometers. If you know acceleration as a vector as a function of time you can easily work out relative speed and distance . Given that when your foot is on the ground its velocity is zero this allows you to work out absolute speed and absolute distance. In the real world most foot pods don't use good enough accelerometers to resolve acceleration as a vector so things get a lot messier. They get around this by requiring you to 'calibrate' the foot pod by travelling a set distance. Garmin for example recommends 800m. It then uses algorithms to sort through the noise and estimate the spatial component of acceleration. This calibration shouldn't be speed dependent but only depends on where you place the sensor. Its not counting strides but trying to work out the horizontal component of acceleration. By knowing an absolute distance it can try and pick out the parts of the accelerometer signal relevant to forward motion and apply it to any situation. How well this works depends on a lot a factors including how good the sensor is and how consistent your stride is. If your stride changes a lot as function of speed you may begin to see a lot of speed related variability.

So to get the distance from acceleration you would have to double integrate. I would think integrating the error of the accelerometer twice would compound and lead to inacurrate results. You seem to be saying there is some magic sauce in their algorithms that compensate for the error through calibration, I wonder how that is done?

At the most basic level, the footpod can function as a pedometer (the way cycling footpods work as cadence sensors). It knows you took 400 steps to go 800m (for example) and can then convert your stride rate into speed, dist, etc. That of course only works if your stride length stays exactly the same. Since stride changes with speed, they have a fudge factor (table) based on averages. Ideally they would take the GPS data from your outdoor runs at different paces and build a curve that was specific to you (I don't see any evidence of this).

Really high quality sensors and algos should, in theory, not need any calibration; since they are measuring all the accelerations in every plane. That is why Stryd is pretty good right out of the box....and also why it is ungodly expensive.

---

ECMGN Therapy Silicon Valley:
Depression, Neurocognitive problems, Dementias (Testing and Evaluation), Trauma and PTSD, Traumatic Brain Injury (TBI)

This is probably the best answer so far. I don't think consumers are really going to know what specific triaxial accelerometer is being used.....or if we did if that information is even useful. We won't know the algorithms either because they are proprietary. Then there's the discussion of how great a location a single foot is for accurately measuring acceleration (and deceleration) of the entire body. It's not used at all for research measuring general physical activity.

It's super easy to break open Stryd, Garmin's pod, or Polar's stride sensor and look at the IMU that is in use. It's typically printed on the silicon. There are dozens of 3-, 6-, 7-, 9-, and 10-Degrees of Freedom IMUs on the market that you can order and program today for less than $20 USD. Adafruit and Sparkfun both sell hobby parts. You can get an Arduino nano and have a working footpod prototype in a week following online tutorials. No, making a working footpod is not that trivial, however the concepts are fairly pedestrian.

The thing that makes Stryd better than most is that is uses a 7-DOF IMU reading at 60hz (I think) in the old version and 10-DOF in Stryd Wind. Or at least they were touting "NEW MAGNETOMETER" in the Stryd Wind marketing. With 7-DOF, you get linear X, Y, Z (from accelerometers), roll, yaw, pitch (from gyroscopes), and barometric pressure sensing. This means you can draw a curvy line in 3D space from all of the measurements. 10-DOF IMUs add in a magnetometer for additional vector corrections based on Earth's magnetic field.

Give the tech a few more years. Stryd shouldn't be the only company selling better measurement accuracy when it comes to footpods.

Breaking it open is certainly a way to find out!

That's a great explanation and also ties in the gyroscope functions (angular velocity) measurements as well. We should probably be saying inertial measurement unit (IMU) instead of accelerometer when referring to the sensors like Stryd, so thank you for adding that. I'm fairly sure 'basic' footpods like Garmin are just accelerometer based and don't add in gyroscopes. I could be wrong. They are most definitely not pedometers though (which you didn't suggest in your post).