
The line over here would describe the data, ignoring the outlier.

But it will give us a huge least squares error over here,

which has an enormous effect on the line.

This one over here makes kind of no sense,

because even though it pulls a little bit on the right side,

it pulled the left side down, and gives a huge least squares error over here.

So that really leaves this one as the most plausible line.

And in reality, you might get something just like this.

With a totally square over here is balanced with many of

these squares errors on this side over here