The reservoir of the M oilfield in Tarim Basin is an unconventional fracture-cave carbonate rock, encompassing various reservoir types like fractured, fracture-cave, and cave, exhibiting significant spatial heterogeneity. Despite the limited pore space in fractures, they can serve as seepage pathways, complicating the connectivity between reservoirs. High-precision fracture prediction is critical for the effective development of these reservoirs. The conventional post-stack seismic attribute-based approach, however, is limited in its ability to detect small-scale fractures. To address this limitation, a novel pre-stack fracture prediction method based on azimuthal Young’s modulus ellipse fitting is introduced. Offset Vector Tile (OVT) gather is utilized, providing comprehensive information on azimuth and offset. Through analyzing azimuthal anisotropies, such as travel time, amplitude, and elastic parameters, smaller-scale fractures can be detected. First, the original OVT gather data are preprocessed to enhance the signal-to-noise ratio. Subsequently, these data are partially stacked based on different azimuths and offsets. On this basis, pre-stack inversion is carried out for each azimuth to obtain the Young’s modulus in each direction, and, finally, the ellipse fitting algorithm is used to obtain the orientation of the long axis of the ellipse and the ellipticity, indicating the fracture orientation and density, respectively. The fracture prediction results are consistent with the geological structural features and fault development patterns of the block, demonstrating good agreement with the imaging logging interpretations. Furthermore, the results align with the production dynamics observed in the production wells within the block. This alignment confirms the high accuracy of the method and underscores its significance in providing a robust foundation for reservoir connectivity studies and well deployment decisions in this region.