Performance on verbal and mathematical tasks is enhanced when participants shift from using algorithms to retrieving information directly from memory (Siegler, 1988a). However, it is unknown whether a shift to retrieval is involved in dynamic spatial skill acquisition. For example, do athletes mentally extrapolate the trajectory of the ball, or do they retrieve the future location from memory? To examine this question, 2 experiments were conducted using a task paradigm similar to the game Pong--a ball was launched from 1 side of the screen and participants attempted to position a paddle to intercept the ball. In Experiment 1, participants responded to a limited number of repeated trajectories. During the learning phase, the response deadline was near the paddle. During the difficult phase, the response deadline was closer to the launch point. During the critical phase, novel trajectories were introduced at the difficult response deadline. If participants are using a retrieval strategy by the critical phase, performance should be significantly worse on the novel trajectories, whereas if they are using an algorithmic strategy, performance on the novel trials should be similar to performance on the repeated trajectories. In Experiment 2, half the participants followed an experimental paradigm similar to Experiment 1 and half experienced all novel trajectories throughout the task. Our results were consistent with a shift from algorithmic processing to retrieval--participants performed significantly better on repeated trajectories relative to novel trajectories. Furthermore, retrieval strategies enhance performance above and beyond what is gained by practicing the algorithm alone.