Suppose that we create another model that simply predicts the average outcome regardless of the value for spread. Does this new model minimize the least squared error? why or why not?.