Cross-validation
Also: k-fold cross-validation, k-fold, k-fold CV, LOOCV
Each fold serves as the validation set exactly once while the others are used for training, and the k scores are averaged. This gives a far more stable estimate than one split and is the standard way to tune hyperparameters on limited data.
CV score = (1/k) Σᵢ scoreᵢ
Example
5-fold CV on 1,000 examples trains five models on 800 points each and scores each on its held-out 200: scores 0.81, 0.84, 0.79, 0.83, 0.82 average to 0.818.