The azimuthal decorrelation between a vector boson and a jet is an essential hard probe in high energy proton-proton and heavy-ion collisions. It suffers from large logarithms in the back-to-back limit, which can be resummed using Soft-Collinear Effective Theory. In this talk I will demonstrate that a suitable choice of jet recombination, the Winner-Takes-All recombination scheme, simplifies the observable tremendously and allows us to derive resummed predictions at next-to-next-to-leading logarithmic accuracy, while maintaining a rich selection of theoretically interesting features. We find for example that linearly polarised collinear functions must be included, and that the observable can be measured with minimal impact on particle tracks. Finally, I will contrast the azimuthal decorrelation with a closely related observable, the radial decorrelation, to motivate why the former is so theoretically accessible, and will briefly comment on the potential impact of factorisation breaking effects and possible extensions to multi-jet decorrelation.