In this note, we provide an elementary analysis of the prediction error ridge regression with random design. The proof is short and self-contained. particular, it bypasses use Rudelson’s deviation inequality for covariance matrices, through a combination exchangeability arguments, matrix perturbation operator convexity.