Pushover analysis is among the most popular of approaches for the seismic assessment of reinforced concrete (RC) structures. In lumped-plasticity structural models, the underlying idea is to approximate the global inelastic behaviour by means of the formation of localized plastic hinges at the frame elements’ ends throughout the incremental nonlinear static analysis. Hence, the proper evaluation of the plastic hinge’s characteristics is essential to obtain reliable results. It should be highlighted in this regard that, even if the plastic hinge region has a clear physical meaning, the notion of plastic hinge length poses some practical problems because of its purely conventional definition. While almost all the existing data-driven proposals for the estimation of such a parameter were carried out by means of a standard linear regression, this study exploits an advanced nonlinear global stepwise regression. Specifically, the evolutionary polynomial regression technique is adopted to develop several formulations for the numerical evaluation of the plastic hinge length. The procedure combines the best features of the numerical regression with those of the evolutionary computing within a multi-objective framework. The formulae that provide the best compromise between accuracy and complexity are selected and then compared with several existing proposals. Pushover analyses also have been performed for two multi-storey reinforced concrete framed buildings by employing different plastic hinge length formulations. The statistical comparative analysis of the results obtained demonstrates the enhanced predictive capability of the formulae calibrated using the evolutionary polynomial regression technique. Moreover, the final numerical applications highlight the role of the plastic hinge length formulation in the seismic assessment of reinforced concrete structures by means of nonlinear static analysis.