Bandwidth smoothing can effectively improve network bandwidth utilization by reducing the burstiness of VBR traffics. The main idea is to transmit the video data of big frames to client buffer prior to their due time. PCRTT can result in a suboptimal transmission schedule with lower computational complexity. However, it is difficult to obtain good-performance parameters, including segmentation and rate computation. In this paper, we first analyze the key difference between the classic smoothing algorithms that can all achieve some optimal metrics. Then we propose an adaptive segmentation algorithm of PCRTT called t-PCRTT according to the fluctuation trend of the playback traffic around the average rate line. As a result, the algorithm can significantly decrease the required client buffer size with the equal number of segments due to its tightly tracking the playback data curve. Using the analysis function, we verify that the required buffer size of t-PCRTT is less than half of PCRTT. The final simulation results using the MPEG4 and H.264 video traces show that t-PCRTT outperforms PCRTT with the same number of segments, and that t-PCRTT almost achieves the optimal number of rate changes and well supports the VCR operation based on VCR-window under various buffer sizes.