When a model learns the noise in the training data instead of the underlying pattern. An overfitted trading model will perform very well on historical (training) data but poorly on new market data. Symptoms include very low error on the training set but high error on a validation/test set. Overfitting happens if the model is too complex or trained too long. To prevent it, traders use regularization, early stopping, or more data. Detecting overfitting involves checking that training accuracy >> validation accuracy, e.g. via k-fold cross-validation.

Back to Robots Glossary Index