We introduce a novel dataset of large depreciations worldwide since 1971. First, we use a multi-step approach to accurately pinpoint large depreciation events on monthly data. We then construct large depreciation episodes that cover 24 months after the initial depreciation event. The month-level data allows for a granular characterization of the dynamics of large depreciations across multiple metrics, including maximum and equilibrium depreciations, overshooting, and number of depreciation events within a single episode. We then present stylized facts on episodes across various characteristics and groups (income, REER trajectory, number of events, exchange rate flexibility, and IMF-supported program status). Among these: (i) a few months into an episode, REER tends to appreciate unless there is an “aftershock” depreciation, (ii) attempts to peg following the initial depreciation are associated with a higher likelihood of “aftershocks”, and (iii) equilibrium REER depreciations are largest when an IMF-supported program is put in place after the initial depreciation takes place.