Just thinking, by integrating changes in X,Y and Z accelerometer values you could theoretically get info about rotational speed.
Ideally, one of them (the one pointing downwards) would not change much except for changes in inclination.
The other 2, let's say X and Y, would alternate maximum and minimum values as the frisbee rotates. The time between 2 maximums in a same axis would theoretically be the rotation period
The main doubt is if they are updated fast enough so that the app can really get all these changes
.
I would start by getting raw data making the phone spin around its center and look for some pattern regarding maximum/minimum values
-EDIT-
I think that what I said about alternating values doesn't make much sense. Anyhow it would be interesting to observe their behavior in a controlled experiment