I was using the term acceleration in it's physics sense "a change in the velocity vector" which means both speed and direction. Crazy meant multiple rapid changes in direction and speed, like say, falling off a bike would result in this "crazy acceleration" as would being hit by something.
Also, sometimes the reaction/result of aggressive drivers isn't catastrophic. For instance, today on my way home from work I had 2 different drivers decide to come within a foot of my bike while passing. This is a calculated move, intended to scare me and teach me a lesson. I know because in one case, I asked when we both got to the same red light, and in the other, the rude gestures of the driver. An accelerometer would have shown nothing, as any reaction on my part would probably have been worse than just hoping the driver didn't judge wrong.
It's not just "tight driving" either, other times drivers are rude, make threatening actions with their vehicles or just plain block my motion preventing me from changing lanes, making turns, and so on, which should be captured, but don't cause anything particularly noticeable in accelerometer data.
Alternately, the false positive rate is probably pretty high here too. I have at least 2 places on my daily commute where I short-cut thru alleys and make some pretty jarring curb passages. There is one particularly tricky transition from street to bike path involving 3 turns in a very short distance (30ish feet). On fast days I bet it looks more like reacting to an aggressive driver than most aggressive driver reactions (from an accelerometer point of view).