"Complete garbage" is something you throw in the trash. I'm interested in the hows and whys of model performance so I can scrape any useful information from them and so that the programmers can improve them. The CFSv2 seems to have some luck in forecasting large scale subsidence and uplift anomalies associated with SST anomalies and general atmospheric dynamics. However, its forecasts of sea ice extent are pretty close to garbage. I'm not satisfied with any of the means that are being used to forecast the melt season. There is no model that does a good job of coupling the physics of the atmosphere, the ice and the ocean to predict sea ice extent, area and volume.
Perhaps the Canadian model does a good job of handling the physics of snow and ice. The American models certainly don't.
The closest year to 2018, in terms of the atmospheric dynamics of NH early spring and late winter, is 2006. Both years had major stratospheric warmings in mid to late February and cold weather that followed in eastern north America and western Europe. Summer 2006 was a relatively good one for preserving Arctic sea ice. Will the Arctic weather this summer be like summer 2006? I suspect not, because of differences in ocean dynamics and ocean heat distribution, but 2006 may be the best analog year to 2018. Tropical Tidbits looks at analog years for helping with seasonal hurricane forecasts. Last year, Levi Cohen's Atlantic hurricane forecast was for less than normal hurricane activity so, be careful, but perhaps there's a 12 year cycle given that 2005 was one of the most active hurricane seasons ever and so was 2017. This is relevant to the Arctic because there's a connection between the ocean heat in the tropical Atlantic and the weather in the Arctic.