The similarity of chromatographic fingerprints is one of the effective approaches evaluating the quality stability of Chinese medicine, and the cosine of angle plays an important role in the application of similarity. However, the cosine approach is insensitive to the data difference when the distribution range of the data sets is wide. When the data proportion of the reference sample and the test sample is greatly different, it confirms that the sensitivity of the cosine to the differences of the peaks owned by both the reference sample and the test sample differs from the peaks owned only by the reference sample or the test sample in this study. The method considers the peaks owned by one sample in addition to peaks owned by both samples, and determines their own appropriate weigh targeting for the maximal homostasis value of proportion among the peaks of all of Smilax glabra Roxb. samples. The method based on sample data could reflect the difference in the chemical composition area ratio between the reference sample and test samples sensitively, and measures the similarity among the nine Smilax glabra Roxb. samples, which is a new similarity algorithm for evaluating the quality stability of herbal medicines.