In the machine learning world, this is the equivalent of overfitting to the training dataset. You can have a model overoptimized to the data used for training a

In the machine learning world, this is the equivalent of overfitting to the training dataset. You can have a model overoptimized to the data used for training and it craters when used in production because the production data has maybe drifted over time, or the training set was not representative for a range of reasons. This is why early stopping is often a good idea when training models, rather than getting the most optimal model for the data being trained on.