This paper analyzes the results of the 2014 SIMCE Mathematics test applied to the 2nd year of secondary education in Chile using Multiple Linear Regression models, emphasizing structural, intermediate and individual variables.Partial and Semi-partial Pearson correlations are reported, informing the interaction between the students' scores (PJE) and variables such as the favorable attitudinal profile towards Mathematics (PREDISP) and the educational quality of the peer effect (INSUF).The RLM analyzes report significant relationships between all the variables of the model and the score, mainly the contributions of INSUF and PREDISP to the variability explained, persistent even controlling the interaction by the rest of the predictors.When segmenting the results by sex, it is evident that the feminization of the course slightly benefits the performance of the males in the group.It is concluded that the performance of this type of tests intervenes skill and predisposition towards learning, which are strengthened or limited by virtue of the educational quality of the influence of peers and social composition that students have.