Drag a line through the scatter to minimize the squared error, then stake your fit by betting where the next point will fall.
In the fit phase, drag the trend line's ends or body through a cloud of points; red residual sticks and a live squared-error readout show how good your fit is. Lock it in, then drag a marker up or down the forecast column to bet where the next hidden point lands. The true point is revealed and scored, and a line that overfits the scatter forecasts worse.
Line of Best Guess is a regression game about least-squares fitting, residuals and the danger of overfitting.
Dragging a trend line through a cloud while red residual sticks and a live squared-error readout respond is least squares: the method that picks the best fit by minimising the sum of the squared residuals.
Drawing the residuals as visible sticks is the original device.
Squaring is the part that needs showing rather than stating. It is why one distant point pulls the line so hard — its residual counts by its square — and dragging the line while the readout moves makes that penalty something you can feel instead of a formula you accept.
Line of Best Guess is a regression game about least-squares fitting, residuals and the danger of overfitting. It's an original free browser game made for PlayPendium.