Computer-aided cooling curve thermal analysis for the casting process is usually applied to predict the latent heat of transformation and the solid fraction. In this study, one of the important limitations of the Newtonian thermal analysis for the prediction of the latent heat was determined as the disregard of the thermal capacity of mold. Therefore, a new baseline technique, two-capacitive system baseline (TCSBL), was developed by taking the thermal capacity of the mold into account. This new method was modeled by considering the analogy between RC circuits and the metal-mold thermal capacitance system. It was improved by a Taylor series expansion approach to express the cooling rates in terms of the metal and mold temperatures. Ten experiments with four types of pure metals were undertaken to compare the latent heat results of TCSBL with the Newtonian and Dynamic Baselines (NBL and DBL). The mean percentage error for the latent heat prediction of TCSBL, NBL, and DBL were calculated as 4.3, 29, and 24 %, respectively, in comparison with the literal value of the latent heats.