Generalization Error

Generalization error is the difference between a model's performance on its training data and its performance on new, unseen data. A high generalization error indicates that the model has failed to learn the underlying patterns and has instead memorized the training data, a classic symptom of overfitting.

In finance, this error is the primary reason why many models fail when deployed. Minimizing this error is the goal of all predictive modeling.

It is achieved through techniques like regularization, cross-validation, and the use of simple, interpretable models. Understanding this error is essential for managing model risk.

It forces the developer to focus on the signal, not the noise. It is the ultimate test of a model's utility.

A model that cannot generalize is essentially useless for real-world trading.