Student grade prediction is a popular task for learning analytics, given grades are the traditional form of student performance. However, no matter the learning environment, student background, or domain content, there are things in common across most experiences in learning. In most previous machine learning models, previous grades are considered the strongest prognosis of future performance. Few works consider the breadth of instructor features, despite the evidence that a great instructor could change the course of a student's future. We strive to determine the true impact of an instructor by analyzing student data from an undergraduate program and measuring the importance of instructor-related features in comparison with other feature types that may affect state-of-the-art student grade prediction models. We show that adding extensive instructor-related features improves grade prediction, when using the best supervised learning classifier and regressor. [For the full proceedings, see ED623995.]