If we know the proportion of the total documents that have been ranked (s) and the number of documents found that contain all the query words (C), we can simply estimate the result size as C/s, which assumes that the documents containing all the words are distributed uniformly