In this paper we present statistical algorithms to classify the stability of proteins by their sequence. A protein sequence consists of successive amino acid codes and can be considered as multivariate categorical data. Based on the statistical variance analysis for data set in each group (stable or unstable protein), the weights are calculated and become an important clue for the effects of the combination of amino acids codes on protein stability. Once the weights for every combination of amino acid codes have been decided, we can assign each protein a score presenting its stability. The distribution of the score for a stable protein is different from the score of an unstable protein. Our algorithm is well suit in the protein stability analysis by its sequence. We propose weighting algorithms and compare them as the results of protein stability classification. It provides an alternative for the protein stability classification and a predictable result as the reference before the protein mutation.