When you click, you're clicking where you saw the object stop. That position already incorporates the brain's forward extrapolation — the model was running ahead of the signal while the object was moving, and when the signal cut off, the model didn't stop at the same instant the object did. The experience of "last seen position" is not a snapshot; it's a model output.
The original Freyd & Finke (1984) paradigm was different: they showed a tilted rectangle, then probed with a second rectangle at the same angle, slightly rotated forward (in the implied direction of motion), or slightly rotated backward. Subjects judged the forward-rotated probe as "same" — their remembered orientation had already moved. The demo here uses position instead of orientation, but the mechanism is the same: the representation includes momentum.
In decelerating mode, the bias should be reduced or absent (Hubbard 1995): the brain's model of a slowing object doesn't extrapolate as far forward. In reversing mode, the last fraction of the trajectory runs backward, and the forward bias should follow the final direction, not the dominant earlier direction. Both conditions show the effect tracks not just velocity but implied trajectory.
What the simulation cannot show: the experience of clicking. From inside, you're clicking where you saw it stop. The overshoot has no phenomenal marker — there is no sense of "I'm clicking slightly past where it was." The discrepancy is only visible from outside, by comparing your click to the ground truth. The simulation provides the outside view; the experience is entirely inside.