Why are ensemble methods superior to individual models?

Ensembled methods

They average out biases, reduce variance, and are less likely to overfit.

There’s a common line in machine learning which is: “ensemble and get 2%.”

This implies that you can build your models as usual and typically expect a small performance boost from ensembling.