Interest in end-of-year accountability exams has increased dramatically since the passing of the NCLB law in 2001. With this increased interest comes a desire to use student data collected throughout the year to estimate student proficiency and predict how well they will perform on end-of-year exams. In this paper we use student performance on the Assistment System, an on-line mathematics tutor, to show that replacing percent correct with an Item Response Theory (IRT) estimate of student proficiency leads to better fitting prediction models. In addition, other tutor performance metrics are used to further increase prediction accuracy. Finally we calculate prediction error bounds to attain an absolute measure to which our models can be compared.