We explore the relationship between a machine-learned structural quantity (softness) and excess entropy in simulations of supercooled liquids. Excess entropy is known to scale well the dynamical properties of liquids, but this quasi-universal scaling is known to breakdown in supercooled and glassy regimes. Using numerical simulations, we test whether a local form of the excess entropy can lead to predictions similar to those made by softness, such as the strong correlation with particles’ tendency to rearrange. In addition, we explore leveraging softness to compute excess entropy in the traditional fashion over softness groupings. Our results show that the excess entropy computed over softness-binned groupings is correlated with activation barriers to rearrangement.