Internal combustion engine modeling is nowadays a widely employed tool for modern engine development. Zero and mono dimensional models of the intake and exhaust systems, combined with multi-zone combustion models, proved to be reliable enough for the accurate evaluation of in-cylinder pressure, which in turn allow the estimation of the engine performance in terms of indicated mean effective pressure (IMEP). In order to evaluate the net engine output, both the torque dissipation due to friction and the energy drawn by accessories must be taken into consideration, hence a model for the friction mean effective pressure (FMEP) evaluation is needed. One of the most used models accounts for engine speed dependent friction by means of a quadratic law, while the effect of engine load (i.e. the thrust that the gas exercises on the piston surface) is considered by means of a linear dependence from the maximum in-cylinder pressure: hence the model requires the calibration of four constants by means of experimental data. The author, on the basis of data acquired during an extensive experimental campaign carried out on the engine test bed, found this model to give an unsatisfying prediction, above all for retarded pressure cycles (i.e. with peak pressure positions higher than 20 crank angle degrees after top dead centre): hence, by means of analysis performed using these experimental data, the author arrived at a new formulation of the friction model, which substantially take into account the effect of engine load by means of the Location of Pressure Peak (LPP). The new model, once calibrated, proved to be effectively more accurate in the prediction of the FMEP than the Chen-Flynn model.