The opposite problem of overfitting, where a model is too simple or not trained enough. An underfitted model fails to capture the market’s structure and performs poorly even on training data. It indicates high bias and low variance. For instance, a linear model that can’t follow a non-linear price trend underfits. One sees underfitting when both training and test error are high. Avoiding underfitting involves using a more expressive model or adding relevant features.

Back to Robots Glossary Index