The aim of this study is to evaluate how far Lempel-Ziv complexity (LZC) of the binary symbolic sequences resulting from static or dynamic transformation (partitioning first-order difference) of the short-term electrocardiogram (ECG) signals (only 2 seconds duration) has the potential in discriminating normal and ventricular tachycardia/ventricular fibrillation (VT/VF) subjects. The statistical analyses show that LZC from either transformation is sufficient to distinguish between normal and VT/VF subjects. Between the two, LZC of dynamic transformation is found to outperform LZC of static transformation. The receiver operating characteristic curve analysis confirms the robustness of this new approach which exhibits an average sensitivity of about 99.1% (100.0%), specificity of about 100.0% (100.0%), precision of around 98.9% (100.0%), and accuracy of about 99.5% (100.0%), with LZCto distinguish between normal and VT (VF) subjects. The presented method is simple, computationally efficient, and well suited for real time implementation in automatic external or implantable cardioverter-defibrillators.