We analyze the worst episodes of currency carry loss in recent decades, including causes, attribution by currency, timing, and duration of carry drawdowns. To explore the determinants of the length of carry losses, we estimate a model of carry drawdown duration. We find evidence that drawdown duration varies systematically with (1) expected return on the carry trade at the onset of the drawdown, (2) financial stress indicators, and (3) the magnitude of deviations from a fundamental value portfolio of the carry-related portfolio holdings. In an out-of-sample test, we show that these determinants can be used to control carry-related losses and improve investment performance.