This letter focuses on the metric sorting step for successive cancellation list decoders for polar codes, which play an important role in the overall comparison process. We propose three effective pairwise metric sorting (PMS) methods based on general pairwise sorting. We first describe a full-sorted PMS architecture with fewer comparators than state-of-the-art methods. Second, we present a half-sorted PMS architecture for notable decreases in the number of comparison stages and augmentation of the comparisons. Finally, we describe the M-bit parallel decoding architecture, which significantly reduces the comparison stages when compared with the 1-bit sorting method of M times. Evaluation results indicate that the proposed PMS method outperforms state-of-the-art metric sorting methods in terms of comparisons and comparison stages, which effectively reduces latency and hardware resource.