Sea ice forecasts are becoming a demanding need since human activities in the Arctic are constantly increasing and this trend is expected to continue. Forecast system development needs to be guided by verification metrics that quantify skill in an appropriate way. Here we apply different verification metrics to real sea ice forecasts to study the behavior of the metrics and to quantify potential predictability, focusing on the sea ice edge position and on subseasonal to seasonal time scales. The employed metrics are the pan‐Arctic sea ice extent (SIE) and area (SIA), the Integrated Ice Edge Error (IIEE), the Spatial Probability Score (SPS), and the Modified Hausdorff Distance (MHD). While the first two metrics evaluate a single integrated quantity, the latter three assess the spatial distribution of the ice cover. Forecasts are verified against the high resolution AMSR‐E and AMSR2 89 GHz sea ice concentration products provided by the University of Bremen. Sea ice forecast products from various research institutes and operational centers are analyzed, in particular those collected within the Sub‐ Seasonal to Seasonal Prediction Project. The analysis covers a time period of 15 years, from June 2002 to June 2017. The forecast systems are characterized by quite different features with regard to the spatial resolution and the complexity of the forecast model, the number of ensemble members and the forecast length. The broad pool of models allows a comprehensive analysis of the metrics’ behavior in different situations, highlighting strengths and weaknesses of the models and of the metrics themselves.