A numerical model to predict static friction for metallic point contacts was developed and validated by comparison to experimental measurements using a specially designed test rig. Key aspects of the numerical model were the incorporation of a digitized real rough surface profile, application of discrete convolution fast Fourier transform (DC-FFT) to predict local asperity interference, and modification of the yield strength to capture the effect of cold hardening. It was found that these model features are critically important to quantitative prediction of static friction. The model significantly underestimated the static friction coefficient if randomly generated surfaces having statistical parameters the same as the measured rough surface were used; digitized real rough surfaces enabled accurate predictions. Further, the model was able to describe the static friction of worn surfaces after cold hardening was introduced through modification of material yield strength. This work illustrates the importance of incorporating the surface features and the change of those features with wear to accurately and reliably predict static friction.