Best Fit Line
Given a series of points that represent some measures in the real world, for example, the heights vs the age of a group of children, when we plot it on a rectangular grid, we get a scatter plot, not necessarily aligned as to form a specific graph.
Sometimes is vital to have a known function that best represents the whole dataset, so we can make predictions.
The best fit line is widely used to do that. We can recognize the best fit line because it usually goes through the maximum number of points and the sum of the distances to each one of them is minimum.
From the three graphs given in the question, it's evident the first line is totally outside the set of points, so it does not do a good job representing the dataset.
The second graph approximates much better to a best-fit line but we can still see the trend of the line does not follow the trend of the dotted path.
Finally, the third graph looks very adequate to represent the scattered points because it keeps always close to them and follows their trend.