The next step would be to start working it without the prop (box jump) in front of the box. Moving to this step before Dancer was doing a good turn 8 out of 10 times was risky: many trainers would keep trying other box prop configurations to force the good turn to happen, instead of removing the prop completely. But for some reason, elaborate box prop configurations really bother me. I mean, reeeeaaaalllly bother me, so I didn't really entertain that option. Recognizing that I might be now settling for double-hitting and/or trapping the ball (I knew I could ensure 4 paws no matter what), I removed the box jump completely.
And what did Dancer do? Five out of ten good turns! Fifty percent! Whoot!
Of course, that means there were 4 out of 10 like this...bad:
And even one that was truly awful:
But 50% is 50%, and I'm pleased! Yes, there's room for improvement (she's coming off a little wide in the good ones, for example). Yes, we could see regression next time (but I doubt it--each time she did a bad one, she did not get rewarded, and then she'd give me a good one after that...as you can see in the video below.)
Now we'll shoot for 8 out of 10 good ones, and then I'll gradually move her further back to about 10 feet from the box, and then I'll start changing my body position (right now, I'm standing on her left side, but in the end, I want to be releasing her from between my feet, with me standing over her.) And then we'll introduce a jump into this whole mess. Yay, progress!!
Here's the video of this session, for anyone who wants to see the good mixed with the bad.