Robotics competitions for K-12 students are popular, but are students really learning and improving their STEM scores through robotics competitions? If not, why not? If they are, how much more effective is learning through competitions than traditional classes? Is there room for improvement? What is the best robotics competition model to maximize students STEM learning? One robotics competition designed to promote the use of math and science is Robofest. Robofest is an autonomous robotics competition with many unique features for STEM education. An example is that students need to solve unknown problems on the day of the competition. The Robofest competition requires use of mathematics and sensors which discourages dead reckoning. Results from 4th-12th graders who completed a STEM assessment before and after the 2010-2011 Robofest competition found students in the Robofest group showed significant improvement in math scores after the competition, compared to students in a control group. These results suggest robotics competitions modeled after Robofest have the potential to maximize STEM learning.