scikit-learn's ready-made baseline. With strategy='most_frequent' it always predicts the most common label.

ExampleDummyClassifier(strategy='most_frequent') scores 0.735 on the test set.