Bias-Variance Trade-off and Overlearning in Dynamic Decision Problems