Making good decisions about whether to commit-to-phase 3 clinical trials is challenging. This is especially true in oncology because the relationships between the registration endpoint, overall survival, and endpoints such as progression-free survival and confirmed objective response are often poorly understood. We present a framework for decision-making based on a three-endpoint win odds. We discuss properties of the win odds and suggest that it can be interpreted, for decision-making, as the reciprocal of an average hazard ratio for overall survival. We confirm the performance of the decision-making method using simulation studies and a clinical trial case study. As part of this work, we describe the simulation of correlated patient-level oncology endpoints using a multi-state model of disease. This model can provide clinically realistic data for testing the performance of analysis methods. We conclude that the win odds can improve commit-to-phase-3 decision-making compared with other methods.