Home Statistics and Visualization Win-Vector Blog When Cross-Validation is More Powerful than Regularization