Individual test students are going to be hard to predict anyway. Though I think this is compared to the sample standard deviation, and for most students their improvement over time is going to be smaller than the difference between two random students.