What is adjusted R-squared and why is it needed?

Adjusted R² penalizes adding unnecessary features.

Unlike R², adjusted R² accounts for number of predictors. It decreases if a new variable does not improve model sufficiently, preventing overfitting through blind feature addition.