The structure of collaboration is known to be of great importance for the success of scientific endeavors. In particular, various types of social capital employed in co-authored work and projects bridging disciplinary boundaries have attracted researchers' interest. Almost all previous studies, however, use samples with an inherent survivor bias, i.e., they focus on papers that have already been published. In contrast, our article examines the chances for getting a working paper published by using a unique dataset of 245,000 papers uploaded to arXiv. ArXiv is a popular preprint platform in Physics which allows us to construct a co-authorship network from which we can derive different types of social capital and interdisciplinary teamwork. To emphasize the 'normal case' of community-specific standards of excellence, we assess publications in Physics' high impact journals as success. Utilizing multilevel event history models, our results reveal that already a moderate number of persistent collaborations spanning at least two years is the most important social antecedent of getting a manuscript published successfully. In contrast, inter- and subdisciplinary collaborations decrease the probability of publishing in an eminent journal in Physics, which can only partially be mitigated by scientists' social capital.