Due to growing concern on brain injury in sport, and the role that helmets could play in preventing brain injury caused by impact, biomechanics researchers and helmet certification organizations are discussing how helmet assessment methods might change to assess helmets based on impact parameters relevant to brain injury. To understand the relationship between kinematic measures and brain strain, we completed hundreds of impacts using a 50th percentile Hybrid III head-neck wearing an ice hockey helmet and input three-dimensional impact kinematics to a finite element brain model called the Simulated Injury Monitor (SIMon) (n = 267). Impacts to the helmet front, back and side included impact speeds from 1.2 to 5.8 ms−1. Linear regression models, compared through multiple regression techniques, calculating adjusted R2 and the F-statistic, determined the most efficient set of kinematics capable of predicting SIMon-computed brain strain, including the cumulative strain damage measure (specifically CSDM-15) and maximum principal strain (MPS). Resultant change in angular velocity, ΔωR, better predicted CSDM-15 and MPS than the current helmet certification metric, peak g, and was the most efficient model for predicting strain, regardless of impact location. In nearly all cases, the best two-variable model included peak resultant angular acceleration, αR, and ΔωR.