Mathematical models for rolling force calculation during hot rolling are crucial for both automatic mill operation and prediction of steel behavior and mill capacity, especially for development of new steel grades and improvement of the existing ones. In literature, the models are based usually on the calculation of steel mean flow stress (MFS), firstly, followed by a force model based on MFS. This approach was applied to a large volume of industrial data of coils produced at Usiminas' plant in Ipatinga, giving unsatisfactory results. Then, the empirical model developed by Schultz was applied to calculate directly the force, with further fine tuning the model with a linear regression taking into account the steel chemical composition. This approach led to a force prediction capability better than the traditional models based on the binomial MFS-force presented in literature, markedly for plain CMn and IF steels. In such cases, more than 90% of force predictions were inside the ±10% error margin, considering the front stands of the hot strip mill.