Most structures working under alternate loadings must be dimensioned to prevent fatigue crack initiation, the main mechanism of mechanical damage in these cases. The various parameters from the fatigue damage prediction models used in these projects should preferably be measured by optimally fitting their equations to well-measured experimental data. In fact, the accuracy of the predictions based on these models depends directly on the quality of the adjustments used to obtain these parameters. As a result, the main purpose of this work is to study the best way to obtain the parameters of the leading prediction models of fatigue crack initiation through experimental data fittings based on the Levenberg-Marquardt algorithm. First, several εN tests were performed on a 6351-T6 aluminum alloy to verify the performance of the proposed fit for the Coffin-Manson and Ramberg-Osgood equations. Then, data from the literature of eight other materials were used to fit classic strainlife models, as well as models based on the Walker exponent, to evaluate the effect of non-zero mean loads in εN tests. Finally, the fitting of an SN model including the Walker exponent was studied, which considers fatigue limits and mean load effects. This study includes as well statistical considerations to quantify the reliability factor from different probability density function assumptions, based on ten data sets from the literature.