In this paper we present an application of Fitness Biasing, a type of Punctuated Anytime Learning, for learning autonomous agents in the space combat game Xpilot. Fitness Biasing was originally developed as a means of linking the model to the actual robot in evolutionary robotics. We use fitness biasing with a standard genetic algorithm to learn control programs for a video game agent in real-time. Xpilot-AI, an Xpilot add-on designed for testing learning systems, is used to evolve the controller in the background while periodic checks in normal game play are used to compensate for errors produced by running the system at a high frame rate. The resultant learned controllers are comparable to our best hand-coded Xpilot-AI bots, display complex behavior that resemble human strategies, and are capable of adapting to a changing enemy in real-time.