The answer is the option A. A best-fit line is close to most of the data point.
A best-fit line intends to minimize the average error (the distance of actual points to the line) , the you might find that the line do not pass through the points, but is close to most of points.