This entry in the 2012 AMS Wind Power Prediction Contest examines the relative influence of changes in the choice of machine learning algorithm and choice of data framework. Algorithms incorporating increased complexity in terms of variable selection and weighting or ensemble methods, such as random forests, neural networks, and multivariate adaptive regression splines, often provide significant decreases in error on a given dataset compared to a simple linear regression method. The magnitude of the improvement is limited by the available variables in the dataset and data quality issues. Deriving additional useful variables from the given data or performing time and space transformations to find more representative data values can also significantly decrease error. This entry evaluates machine learning algorithms with varying complexities on different formulations of the training data with cross validation to determine which combination of algorithm and transformation provide the most improvement. That combination is submitted as the official entry for the contest.