Several sources of error make raw GPS measurements inaccurate. To make up for this, consumer GPS receivers use a "model" to help determine what part of the measurement is error. Basically, the model tells the receiver what kind of motion it can expect. If a measurement indicates motion that doesn't fit the model, then the receiver assumes it must be noise, and ignores it.
In many consumer GPS receivers, the model is presented as a choice between "car" and "pedestrian" modes. In the FlySight, a few more choices are available, and it's not too hard to guess what kind of motion each might assume:
|Pedestrian||Slow, mostly horizontal motion|
|Automotive||Fast, mostly horizontal motion|
|Sea||Entirely horizontal motion|
|Airborne||3D motion with limited acceleration|
It's important to understand the role the model plays. For example, if you set the FlySight to use the "Pedestrian" model, and then you jump from a plane, it will try to reconcile your rapid descent with its idea of how a pedestrian moves.
Less obviously, if you tell the FlySight to use the "Airborne < 1G" model, and then you perform a maneuver which results in more than 1G acceleration, the FlySight will "overshoot", because it chalks some of that acceleration up to noise.
When flying a wingsuit, we recommend the "Airborne < 1G" model. For swoopers, we recommend the "Airborne < 2G" model", since high performance landings can generate more than 1G acceleration.