When using a DLT algorithm to estimate a homography, would using more points result in more or less error?

Question

In page 116 of Multiple View Geometry, the graphs compare DLT error (solid circle), Gold Standard/reprojection error (dash square), theoretical error (dash diamond) methods of estimating a homography between figure a and the original square chess board. Figure b shows that when you use more point correspondences to do such an estimation, there is a higher residual error (see figure below). This doesn't make sense to me intuitively, shouldn't more point correspondences result in a better estimation?

Answer 1

The residual error is the sum of the residuals at each point, so of course it grows with the number of points. However, for an unbiased algorithm such as the Gold Standard one, and a given level of i.i.d. noise, the curve flattens because the contribution of each additional point to the sum counts less and less toward the total.