Recently, many messages are submitted and read using microblog Web services, such as twitter, face book, and others. These messages are written by unspecified users, these messages consists of unreliable contents. Therefore, to specify which contents are reliable or not, we should analyze messages. However, these messages are too short, we are hard extract meaningful, useful features from these messages. In this paper, we propose a method to assess credibility values of messages using remaining ratio of ReTweets. We assume that if a high credibility messages are retweeted, the original messages are remain. However, if a low credibility messages are retweeted, several terms are added about opinion of users. Therefore, if there are many retweets of a message, and these retweets are remain, the message should be credible. Using this assumption, we propose a method to calculate credibility degrees of messages using added and deleted messages of retweets.