Why is a graphical determination of the average velocity obtained from the slope of a best-fit line drawn for the data more accurate than calculating several velocities over small time intervals and averaging them together?