The line of best fit does not have to start from 0.
It is called the line of best fit because it tends to satisfy all the possible points in consideration at the same time with minimal variation.
The line that minimized the sum of the squares of the diffences of each point from the line is the line of best fit.
A line of best-fit.
If most of them lie below the line, then that line isn't the best fit. The exact layout depends on what definition you use for "best fit", but any definition will produce a line that has roughly the same number of data points on each side of it.
Because the "best fit" line is usually required to be a straight line, but the data points are not all on one straight line. (If they were, then the best-fit line would be a real no-brainer.)
The line of best fit is the best possible answer you can get from raw data. They also can be used to make predictions.
Finding the line of best fit is called linear regression.
The line of best fit does not have to pass through the 0 (origin) and rarely does