Consider a principal who assigns a job with two tasks to two identical agents. Monitoring the agents’ efforts is costly. Therefore the principal rewards the agents based on their (noisy) relative outputs. This study addresses the question of whether the principal should evaluate the outputs of each task separately and award two winner prizes, one for each task, or whether it is better to award only one winner prize to the agent who performs the best over the two tasks. There are two countervailing effects. First, there is a prize-diluting effect, because for a given budget, the prizes will be smaller when there are two winner prizes than when there is only one winner prize. The prize-diluting effect reduces the agents’ incentives to invest their effort when there are two winner prizes. Second, there is a noise effect because the noisiness of the evaluation is reduced when there are two winner prizes. The main contribution of this study is to show that the prize-diluting effect dominates the noise effect. Hence, in general, principals will award prizes for combined tasks, and not for separate tasks. Several extensions are considered to test the robustness of this dominance result.