The background of this research is the low mathematics learning outcomes of class X students of SMKN 10 Merangin, because the learning model used by the teacher has not varied and the teacher also pays little attention to the potential in students, namely students' self-confidence. The aim of the research is to describe learning outcomes in terms of self-confidence that it is better to use the Teams Games Tournament (TGT) learning model than conventional learning for class X students of SMKN 10 Merangin for the 2021/2022 academic year. The population of class X students consisted of 7 classes totaling 166 students and the research design was a 2x3 factorial design. Class X TBSM 2 is the experimental class, while class X TBSM 3 is the control class obtained from the sampling method, namely Simple Random Sampling, then the 2 sample classes are grouped based on their self confidence. Data collection techniques through tests using instruments in the form of essay questions and self-confidence questionnaires. The data analysis technique for testing hypotheses 1,2,3 and 4 is the t-test while the fifth hypothesis is a two-way ANOVA. The conclusion of the test is: The TGT learning model performs better than conventional learning on overall student mathematics learning outcomes and moderate self-confidence. Meanwhile, the mathematics learning outcomes of students who have high and low self-confidence are worse or equivalent to conventional learning. Then, according to this research, the TGT learning model and self-confidence do not affect students' mathematics learning outcomes.