The study area is located in the northwest of the Chaixi Depression in the Qaidam Basin， and the sedimentary facies is composed of shallow lake and semi-deep lake interdeposits, with marl, algal limestone, sandstone, shale and mudstone interlayered with thin single-layer thickness. The wells drilled into the target zone are few and unevenly distributed, the petrophysical characteristics are not so sensitive, and the physical properties of the reservoir change quickly, showing strong reservoir anisotropy. All these factors bring the difficulty to predict the sweet spots of complex mixed rock. In order to better predict sweet spots of the reservoirs with complex lithology and strong anisotropy, this study relies on the newly acquired “two-wide and one-high” 3D seismic data, and its spiral trace gathers based on OVT domain migration processing. Firstly, the azimuth superposition data is applied to carry out anisotropy analysis, and the 0° and 90° azimuth are identified as the sensitive azimuth of faults or fractures and petrophysical property respectively. Secondly, on the basis of the structure analysis, the seismic attributes used to describe faults and fractures are extracted from the sensitive orientation. Furthermore, fracture prediction, inversion of the selected carbonate content curve and porosity co-simulation are implemented to characterize the space distribution of reservoir fractures and petrophysical property. Eventually, through comprehensive analysis, the sweet spots of the mixed rock are well predicted, which provides the basis for the next well location deployment.