Depression is clinically characterized by obvious changes in decision making that cause distress and impairment. Though several studies suggest impairments in depressed individuals in single tasks, there has been no systematic investigation of decision making in depression across tasks. We compare participants diagnosed with Major Depressive Disorder (MDD) (n = 64) to healthy controls (n = 64) using a comprehensive battery of nine value-based decision-making tasks which yield ten distinct measures. MDD participants performed worse on punishment (d = -0.54) and reward learning tasks (d = 0.38), expressed more pessimistic predictions regarding winning money in the study (d = -0.47) and were less willing to wait in a persistence task (d = -0.39). Performance on learning, expectation, and persistence tasks each loaded on unique dimensions in a factor analysis and punishment learning and future expectations each accounted for unique variance in predicting depressed status. Decision-making performance alone could predict depressed status out-of-sample with 72% accuracy. The findings are limited to MDD patients ranging between moderate to severe depression and the effects of medication could not be accounted for due to the cross sectional nature of the study design. These results confirm hints from single task studies that depression has the strongest effects on reinforcement learning and expectations about the future. Our results highlight the decision processes that are impacted in major depression, and whose further study could lead to a more detailed computational understanding of distinct facets of this heterogeneous disorder.