The disc is measuring way to far long before the throw vs parsing data from a key point.
The opposite - it's sometimes not recording for long enough.
All i know for sure is that it needs to be stationary to calibrate. Here's what i imagine is happening - a bit of guesswork but i bet it's not far off.
When the disc is stationary, the combined acceleration measured by all the accelerometers is just going to equal gravity (9.81m/s/s). So if it measures exactly that acceleration for some short period of time, it can assume that it's stationary, and the exact measurements on the various accelerometers, at their different angles (i.e. the individual measurements which we combined to make 9.81) can be used to calculate its orientation.
[COLOR=var(--text)]Any acceleration it feels from that point on can then be used to calculate the movement from that stationary point.
With modern accelerometers, i see no reason it can't be extremely accurate. The only issue is if the uploaded data doesn't contain a stationary point. (It records data all the time, but only uploads the 6 seconds or so up to release, so if the acceleration is never 9.81 in that period then it won't have much of a clue what 'stationary' is and you'll get weird results.)
You could try to game it by (e.g.) walking very smoothly backwards at 5mph, so that it thinks it's stationary when it's actually doing
minus 5mph. Your throws will register 5mph faster. The acceleration it feels when moving at constant velocity is the same as it feels when stationary - vis Newton or Einstein - which would be a problem if it was easy for a human to move it at constant velocity, but to do so accurately enough to fool the very sensitive accelerometers is going to be bloody difficult. You're going to wobble about a bit.
So in practice, the only problem is if it's never stationary in the 6 seconds and the software starts trying to find a stationary point that doesn't exist. I'd prefer it to throw up an error message in that situation, but actually it does give some funky results.
But none of those funny results has any impact on its accuracy in 'normal' use.
I think it would be better if they explained how it worked a bit, so that people could understand what not to do with it. I guess they have their reasons not to.
So - you're right that it can do funky things, and i have big doubts about stokeley's 86mph. People will misuse it, sure. But i wouldn't throw the baby out with the bathwater and assume it's not normally accurate. I think it's extremely accurate (in measuring, not the simulator obviously) for normal throwing, and very useful. Chasing data isn't silly in itself, it's only silly to chase the occasional glitchy data. (But as you say, people are pretty dumb sometimes.)[/COLOR]