A theoretical method for predicting the smoke point of pure hydrocarbon liquids is presented. The method is based on a structural group contributions approach and does not require any experimental procedures or information of fuel properties, other than the molecular structure of the fuel molecules. The proposed correlation is presented in the form of a multivariable regression. The average deviation is only 1.3 TSI (threshold soot index) units for similar to 70 compounds from low-sooting paraffins to highly sooting aromatics, and the average relative error is 9.08%. The results of three different sets of structural groups derived from the Quann and Joback group contribution methods are tested and compared. For a mixture with a defined composition, the estimation of smoke point is also discussed. The method is of potential value for the formulation of surrogate fuels of hydrocarbon mixtures, where matching the fuel's sooting tendency is important.