Dark pools are a relatively recent type of equities exchange in which transparency is deliberately limited in order to minimize the market impact of large-volume trades. The success and proliferation of dark pools has also led to a challenging and interesting problem in algorithmic trading --- namely, optimizing the distribution of a large trade over multiple competing dark pools. In this work we formalize this as a problem of multi-venue exploration from censored data, and provide a provably efficient and near-optimal algorithm for its solution. This algorithm and its analysis has much in common with well-studied algorithms for exploration-exploitation in reinforcement learning, and is evaluated on dark pool execution data from a large brokerage. Joint work with Kuzman Ganchev, Yuriy Nevmyvaka, and Jennifer Wortman Vaughan.