Describe the bug

https://scikit-learn.org/stable/auto_examples/calibration/plot_calibration_curve.html

In the calibration curve page, a "scores_df" is generated to showcase supporting model evaluation metrics in addition to the calibration curves.

I noticed that my ROC AUC score was unusually low and noticed that it was consuming Y_PRED instead of Y_PROB. This is incorrect and may confuse users in the future.

Steps/Code to Reproduce

Not applicable.

Expected Results

Expected AUC in the 65-75 range for my application.

Actual Results

Observed AUC in the 50-55 range instead.

Versions

Not really relevant but 1.2.1.