Shouldn't the x domain always be [0, 1]?

You mean for the speed? I'm using a symmetric difference quotient to approximate the speed which results in the endpoints being lopped off. I suppose that an appropriate one-sided difference quotient could be used at either endpoint but it doesn't seem too important really.

I meant for both. If you look at easeBack you can see that both speed and position are plotting points for x<0 and x>1. But I might be misunderstanding what x represents, considering that the sample points aren't evenly spaced.

I didn't even notice that! But, yes, you're right of course. The code I based this on used both an ease and a tween, which (I think) interfered with one another. I've modified it to use just an ease to remove that interference. I also changed it so that the position functions are simply extracted from the 'x' attribute of the rectangle. I don't think there's any reason that the sample points need be evenly spaced. On the contrary, my impression is that the objective is to capture the actual observed position and speed.