Introduction: Saliva is the most promising biofluid to monitor the physiological state of athletes, because this method is not invasive and has low contamination risks. The characterization of saliva by Fourier transform infrared spectroscopy (FT-IR) has been studied as an alternative technique to the standard clinical analysis. However, methodological procedures for saliva analysis are not completely clear, especially in terms of influence of storage conditions and sample preparations for infrared analysis. Thawed saliva includes a precipitate, which may influence the infrared spectral analysis. Thus, the purpose of this study was to show the spectral differences of the precipitate, supernatant, and a combo, as well as the best way to classify the physiological state of the athletes by FT-IR. Methods: The saliva collection was performed before, immediately after, and two hours after a handball match. After the storage of samples at -20 ○ C, it was possible to identify two phases (precipitate and supernatant) and to determine the biochemical differences between the spectra of each phase, which were distinctly analyzed by the second derivative and deconvolution bands. Results: The precipitate and supernatant results showed characteristic bands, especially in the protein regions. All FT-IR spectra were also statistically classified by linear discriminant analysis (LDA), using principal component analysis (PCA). The LDA precipitate and supernatant had lower value when compared to combo spectra (Combination of precipitate and supernatant) with 82%, showing that this combination is the best way to discriminate spectra of saliva samples collected before, immediately after, and 2 h after physical effort. Discussion: The results showed that it is possible to differentiate biochemically the two salivary phases, as well as the importance of the homogenization process of saliva samples to classify the physiological status of athletes using FT-IR.