As an increasing number of users access information on the Web, there is a great opportunity to learn about the users?probable actions in the future from the server logs. In this paper, an n-gram based model is presented to utilize path profiles of users from very large data sets to predict the users?future requests. Since this is a prediction system, the recall cannot be measured in a traditional sense. Therefore, the notion of applicability is presented to give a measure of the ability to predict the next document.The new model is based on a simple extension of existing point-based models for such predictions,but the results show that by sacrificing the applicability somewhat one can gin a great deal in prediction precision.The result can potentially be applied to awide range of applications on the Web,including pre-sending,pre-fetching,enhancement of recommendation systems as well as Web caching po;icies.The tests are based on three realistic Web logs.The new algorithm shows a marked improvement in precision and applicability over previous approaches.