Hi pvareschi,  Here's the reply from the instructor:  Bias would not occur with an overfit model.  Only an underfit model would “systematically” predict values either larger than or smaller than the true target value.  That’s what bias in this sense means. When we talk about a model being “high variance” we mean that it is modeling or capturing the unpredictable and unrepeatable random variation in the data.  All data has such random variation, but only when models are overfit do they try to capture this random variation.  A model that fits the data “just right” is only capturing the predictable and repeatable patterns in the data Best,  theresa ... View more