This shift is represented by a motion vector.

The motion vectors for each 8x8 or larger partition region can point to different reference pictures.

Thus the overall size of motion vector plus prediction error will be greater than the raw encoding.

The interpolated frames (namely B-frames) require more motion vectors which means an increased bit rate.

These modes are referred to when a block is coded without sending residual error or motion vectors.

A possible criterion could be to copy the motion vector from a neighboring block.

P-frames have 1 motion vector per macroblock, relative to the previous anchor frame.

The offset is encoded as a "motion vector."

This data is then stored as a motion vector for this block.

The ball itself has a high motion vector, but its background that makes up the rest of the block might not have any motion.