Whenever you're ready, walk me through how you'd approach this.
You are comparing several supervised learning models and want a principled way to explain why some underfit while others overfit.
Can you define the bias-variance tradeoff and explain how it impacts model performance?