SST = SSM + SSE

Start with the total variation in y, then separate it into the model's part and the remaining error.

Build the line

The points update to keep this the least-squares line.

Squares inside the plot

Choose one or more.

Regression squares

Each colored vertical distance becomes the side of a square.

SST squares: observed y to mean y
-3.5-1.70.01.73.5-9-5059mean yxy
SST = SSM + SSE108.57 = 100.80 + 7.77SST = 108.57

Adjusting the controls regenerates the points, so the shown line remains a least-squares fit with an intercept.