This paper expands the scope of typical quantitative studies on mathematics self-efficacy by including a short ethnography of the students' daily classroom experiences. It attempts to provide a "thicker description" and a context in which these beliefs could be interpreted. Using both qualitative and quantitative sets of data, it is argued that mathematics self-efficacy is not a strong predictor of grades, which was assumed to approximate mathematical performance. In addition, it is hypothesised that it has less influence on behaviour related to mathematics learning than other factors such as "hiya" or anxiety or instructional practices in the classroom. Data for this research were gathered from a rural high school in the Philippines.