The Arctic Ocean is of central importance for the global climate and ecosystem. It is a region undergoing rapid climate change, with a dramatic decrease in sea ice cover over recent decades. Surface advective pathways connect the transport of nutrients, freshwater, carbon and contaminants with their sources and sinks. Pathways of drifting material are deformed under velocity strain, due to atmosphere-ocean-ice coupling. Deformation is largest at fine space- and time-scales and is associated with a loss of potential predictability, analogous to weather often becoming unpredictable as synoptic-scale eddies interact and deform. However, neither satellite observations nor climate model projections resolve fine-scale ocean velocity structure. Here, we use a high-resolution ocean model hindcast and coarser satellite-derived ice velocities, to show: that ensemble-mean pathways within the Transpolar Drift during 2004–14 have large interannual variability and that both saddle-like flow structures and the presence of fine-scale velocity gradients are important for basin-wide connectivity and crossing time, pathway bifurcation, predictability and dispersion.