River avulsions help distribute sediment across the floodplain, and the frequency and flow path selected by each avulsion determines flooding and sedimentation patterns. However, rivers avulse infrequently, making it difficult to quantify flow-path behavior and test hypotheses posed by experimental and numerical studies. We used Google Earth Engine to create a novel data set of 55 avulsions that occurred from A.D. 1984 to 2014 in the Andean and Himalayan foreland basins. On each avulsion we measure hop length (the displacement of the avulsion across the floodplain orthogonal to the parent channel) and avulsion length (the length of the parent channel reach). Without controlling for external factors, such as climate or geology, we find that both hop length and avulsion length scale with parent channel-belt width. Avulsion channels tend to hop 2.5 channel-belt widths away from their parent channel and avulsion length is 13.4 parent channel-belt widths. We use these data to test between end-member scenarios where flow paths are randomly selected or steered by floodplain topography. Observed avulsions are inconsistent with flow-path predictions derived from a directed random walk model, and a scaling analysis of alluvial ridge size shows that hop lengths are set by alluvial ridge widths. These results suggest that avulsion flow-path selection in the Andean and Himalayan basins is driven by alluvial ridge topography, which promotes evenly spaced (or compensational) channel sand bodies over decadal time scales. These data and results are important for understanding controls on avulsion flow-path selection and how avulsion processes are represented in the stratigraphic record of foreland basins.