The purpose of this study was to identify whether there are differences in learning outcomes and dispositions of students who get the game and conventional learning. The population of this study was overall Junior High students grade VII in Semarang. Sampling was done by stratified cluster random sampling. The sample in this study were 200 people with the details of 100 students for the class learning with the game “jirak”, 100 students for the class with conventional learning. Based on the hypothesis test we concluded that: 1) there were differences in learning outcomes and dispositions of students learning with game and conventional (Wilks Lambda = 0439 and p value (sig.) <5%). 2) the relationship between learning model with the results of student learning had a significance level (sig.) <5%, then Ho was rejected. The average of student learning outcomes with the game “jirak” was greater than the average of student learning outcomes with conventional learning, it can be concluded that student learning outcomes with the game “jirak” better than student learning outcomes with conventional learning. 3) the relationship between students' learning model with a disposition has a significance level (sig.) <5%, then Ho was rejected. Because the average disposition of students with learning games larger than the average student disposition to conventional learning, it can be concluded the disposition of students who had learning with the game better than the disposition of students who had conventional learning.