Teachers in Sweden assess and grade their students, without any external examination. The overarching question that is studied in this report is whether teachers can predict the test results of their students. Similar studies, have been made earlier by other researchers in other countries, but only one has focused on teachers’ experience and gender. All the others had the students in focus. The purpose of this report was to study the correspondence between students’ actual test results and the teachers’ prediction of the test results. Another purpose was to study what factors influence the teachers’ ability to predict test results correctly, the teachers’ background, their experience, and their cooperation with colleagues. Data was collected in two steps. First, the teachers predicted their students’ results on the national test in mathematics course B for spring 2004 before the teacher could see the present test. Second, the teachers sent in the result of the national test and answered a questionnaire. The results indicate that the correspondence between the predicted and the actual test result was somewhat higher than in many other studies. Large schools, experience, cooperation with colleagues and more information about the students made the teachers predict their students’ test results more correctly. But which programme the students were studying also influenced the teachers’ ability to predict correctly. It would be interesting to study how teachers think when they predict their students’ test results. Another question for further research is whether teachers’ predicted test result could be a better measure of the students’ test ability than the test score.