Why is the standard error of regression so low using historical data?: Comments on "size matters"