Hip fracture, which is often due to osteoporosis or other conditions affecting bone strength, can lead to permanent disability, pneumonia, pulmonary embolism, and/or death. Great effort has been directed toward developing noninvasive methods for evaluating proximal femoral strength (fracture load), with the goal of assessing fracture risk. Previously, computed tomographic scan-based, linear finite element (FE) models were used to estimate proximal femoral fracture loads ex vivo in two load configurations, one approximating joint loading during single-limb stance and the other simulating impact from a fall. Measured and computed fracture loads were correlated (stance, r=0.867;​ fall, r=0.949). However, precision for the stance configuration was insufficient to identify subjects with below average fracture loads reliably. The present study examined whether, for this configuration, nonlinear FE models could be used to identify these subjects. These models were found to predict fracture load within ±2.0 kN (r=0.962). This level of precision is sufficient to identify 97.5% of femora with fracture loads 1.3 standard deviations below the mean as having below average fracture loads. Accordingly, 20% of subjects with below average fracture loads, i.e. those with the lowest fracture loads and likely to be at greatest risk of fracture, would be correctly identified with at least 97.5% reliability. This FE modeling method will be a powerful tool for studies of hip fracture.