Oil-water discrimination is of great significance to design and adjust development project for oilfields. For fractured reservoir, based on anisotropic S-wave splitting information, it becomes possible to effectively solve such problem that the traditional longitudinal wave exploration is difficult to deal well with due to the similar bulk modulus and density of these two fluids. In this paper, by analyzing the anisotropic character of the Chapman (2009) model, the velocity and reflection coefficient differences between fast and slow S-wave caused by fluid substitution were verified.Then through wave field response analysis of the theoretical model, we found that water saturation causes larger time delay, larger time-delay gradient and lower amplitude difference between fast and slow S-wave, while the oil case is corresponding to lower time delay, lower gradient and higher amplitude difference. Therefore, a new class attribute was proposed, that is, the amplitude energy of the fast and slow shear wave, used for oil-water distinction. This new attribute and the time delay gradient attribute were both applied to the 3D3C seismic data of carbonate fractured reservoirs in the Luojia area of Shengli Oilfield in China. It shows that the prediction result of energy attribute has a better consistency with well information than time delay gradient attribute, and hence demonstrates the great advantages and potential of this new attribute in oil-water recognition.