Concept drift detectors allow learning systems to maintain good accuracy on\nnon-stationary data streams. Financial time series are an instance of\nnon-stationary data streams whose concept drifts (market phases) are so\nimportant to affect investment decisions worldwide. This paper studies how\nconcept drift detectors behave when applied to financial time series. General\nresults are: a) concept drift detectors usually improve the runtime over\ncontinuous learning, b) their computational cost is usually a fraction of the\nlearning and prediction steps of even basic learners, c) it is important to\nstudy concept drift detectors in combination with the learning systems they\nwill operate with, and d) concept drift detectors can be directly applied to\nthe time series of raw financial data and not only to the model's accuracy one.\nMoreover, the study introduces three simple concept drift detectors, tailored\nto financial time series, and shows that two of them can be at least as\neffective as the most sophisticated ones from the state of the art when applied\nto financial time series. Currently submitted to Pattern Recognition\n