Ventricular late potentials (VLP) are low amplitude and high frequency transients registered on high resolution electrocardiogram (HRECG), markers of life threatening ventricular tachyarrhythmia. This study assessed a novel VLP group classification method based on principal components (PC) analysis. Thirty-six subjects (mean ± SD; 55.4 ± 11.6 years) divided in two groups, 18 healthy controls and 18 patients with induced sustained monomorphic ventricular tachycardia were included. Four PC data matrix from HRECG signal averaged with no further filtering leads were built, taking QRS onset as reference. Mahalanobis distance calculation combined with classification by logistic regression determined optimal separation threshold between groups for each matrix. HRECG signals were also analyzed using classical approaches. ROC curve analyzes compared novel and classical methods (α < 0.05). The optimal configuration retained seven initial PCs. Average c-statistic was 0.99 for PC method and 0.65 for classical methods taken together (p < 0.05). PC analysis increases diagnostic accuracy for VLP group classification with potential clinical application.