Current data mining approaches generally condense play information into characteristics or features, which are then analyzed for patterns. This works best when the designer has specific questions that the data can answer. It is more difficult to formulate queries when one is not sure what patterns are in the data. For example, it is difficult to detect player confusion by looking at game metrics, even when it would be obvious by watching a video of the game play. Humans are good at recognizing these kinds of patterns visually, and so it is easier to identify complex patterns by including a person in the analysis [13]. This is the premise of visual data mining, which has been been applied to many kinds of data [14]. It is particularly useful when one does not know beforehand what patterns one will find, such as in games.