Residuals are the difference between an observed or true value and a predicted (fitted) value. For example, in the following diagram, one of the residuals is (A-B), where A is the observed value and B is the fitted value:
The preceding scatter plot depicts that we are fitting a line that could represent the behavior of all the data points. However, one thing that's noticeable is that the line doesn't pass through all of the points. Most of the points are off the line.
The sum and mean of residuals will always be 0. ∑e =0 and mean of e =0.