This plots the swing in RA during tracking after 2 manual PEC trainings. The maximum deviation is now 12 arcseconds, down from the nearly 30 of the untrained tracking! A major improvement, but still pretty jumpy.

Below the error between frames is shown. Aside from the big jump around sequence number 100 (which I believe was wind related), the error between 10 second frames regularly peaks around 4 arcseconds. This is twice as smooth as the error measured with the untrained PEC.