In this paper, we deal with layered predictive compression of animated meshes represented by series of 3D static meshes with same connectivity. We propose two schemes to improve the prediction. First improvement is using weighted spatial prediction rather than averaging neighbor vertices. The second improvement is a novel predictor based on rotation angle of incident triangles in current and previous frames. The experimental results show that around 6- 10 % bitrate reduction can be achieved by replacing the spatial prediction in the reference coder with the proposed weighted spatial prediction and 9- 18 % bitrate reduction is possible with the proposed angle based predictor using weighted spatial prediction, depending on the content and quantization level.