The depth of decision tree represents the degree of how optimal it is. We first need to know that the longest tree leads to a model that learns more complex information i.e. overfitting. On the other hand, shallower trees prone to learn less complex information i.e. underfitting.

Simpler models/trees usually generalize better than the complex ones. In other words, a tree that has a short path from its root performs better decisions than a model having longest path.