The predictive power of a classical molecular dynamics simulation is largely determined by the physical validity of its underlying empirical potential. In the case of high-energy collision cascades, it was recently shown that correctly modeling interactions at short distances is necessary to accurately predict primary damage production. An *ab initio* based framework is introduced for modifying an existing embedded-atom method FeNiCr potential to handle these short-range interactions. Density functional theory is used to calculate the energetics of two atoms approaching each other, embedded in the alloy, and to calculate the equation of state of the alloy as it is compressed. The pairwise terms and the embedding terms of the potential are modified in accordance with the *ab initio* results. Using this reparametrized potential, collision cascades are performed in Ni_{50}Fe_{50}, Ni_{80}Cr_{20} and Ni_{33}Fe_{33}Cr_{33}. The simulations reveal that alloying Ni and NiCr to Fe reduces primary damage production, in agreement with some previous calculations. Alloying Ni and NiFe to Cr does not reduce primary damage production, in contradiction with previous calculations.