The article introduces a new method for testing how well a statistical model fits real data. This method combines hypothesis testing and nonparametric density estimation to create a more informative test. The researchers developed a smooth test that selects the best model based on a specific loss function. When the test rejects the null hypothesis, they recommend plotting the selected model estimate, which minimizes errors. This approach is useful when traditional tests are not effective. The researchers also created a new variance estimator and diagnostic tests to improve the procedure's accuracy. They tested these methods on logistic and extreme value distributions with promising results.