Bias-Variance Trade-Off is a trade-off related to model complexity and the associated types of generalization error. A model which is too simple may not capture the nuances of the data, and make generalization errors because of gross assumptions (bias). On the other hand, a model which is too complex may not be able to ignore the natural randomness present in the training data, leading to predictions that vary a lot depending on the particular realization of the training set.