In this New Zealand study, the maths self-efficacy and achievement of Year 4-10 students is being tracked over 3 years. A particular strength of this study is that maths self-efficacy is being measured in relation to items that have been psychometrically located on a difficulty scale, enabling students' self-efficacy responses to be interpreted in relation to the same scale. The clear picture emerging from an analysis of the first round of data was that, on average, students' expectations of success on a range of mathematics items exceeded what could reasonably be predicted based on their actual achievement. This applied across all students, regardless of their characteristics (gender, ethnic group, year level) and school decile. However, the extent to which students' maths self-efficacy exceeded their predicted achievement varied between groups of students, with those in low-decile schools tending to report self-efficacy levels that exceeded their predicted achievement by more than their peers in mid and high-decile schools. [This research is funded through NZCER's Government Grant.]