Evaluation of the SIDC forecast and models to predict an X-ray flare of classes M or X, across the years (a) and for each individual forecaster (b). The following description holds for both plots. Top panel: POD (lines, left vertical axis, Eq. (5)) and HSS (bars, right vertical axis, Eq. (10)). Middle panel: PC (lines, left vertical axis, Eq. (6)) and the TSS (bars, right vertical axis, Eq. (11)). POD and PC are shown for all forecasts, while HSS and TSS are only for the SIDC forecast. Bottom panel: the proportion of days with an event, a hit, a miss and a false alarm (on a logarithmic scale). The proportion of hits and the proportion of misses sum up to the proportion of events. The number of days on duty, the bias and the difference of the TSS with the TSS of the persistence model are mentioned for each forecaster in the table below the figure. Note that some verification measures can not be calculated and are left undefined in case of lack of events or hits (e.g. year 2009 and forecaster 11).