Even when you have more variance in your Sportband runs (bigger spikes) those spikes also snap to a set of fixed levels. Almost like the snap-to-grid feature in something like Photoshop. Getting distance measurements every 10 seconds from a pedometer/accelerometer is tricky and this artificial limiting of the readings is working against you. It's a bit like asking people to guesstimate the height of a tree. You might get guesses of 30ft, 33ft, 28ft and 27ft. Let's say the actual height of the tree is 29ft. These guesstimates when averaged (actually a least squares regression in our case) gives you 29.5ft.
But, now introduce this limiter that says you can only guess 25, 30 and 35. Those previous 4 guesstimates might now become 30, 35, 30 and 25. In this case we get 30.0ft instead of 29.5 so we are still getting close. But that 27ft guess that we snapped down to 25 could easily have snapped to 30 instead and if it did we would suddenly swing up to an average of 31.25.
The point of all this is that this artificial granularity limitation in the Sportband creates instability. You may get lucky 18 times in a row, but then on the 19th you get a series of 27's turned into 30's in a row and suddenly your overall distance is way out of whack. And I call this an artificial limitation because the sensor that is doing the actual measurement is the same between the Nano and the Sportband. The Sportband just receives the signals from the sensor and records them. I see no reason for it to not record them in the exact same way as the Nano.
