Online pick’em games, such as the recent NCAA college basketball March Madness tournament, form a large and rapidly growing industry. In these games, players make predictions on a tournament bracket that defines which competitors play each other and how they proceed toward a single champion. Throughout the course of the tournament, players monitor the brackets to track progress and to compare predictions made by multiple players. This is often a complex sensemaking task. The classic bracket visualization was designed for use on paper and utilizes an incrementally additive system in which the winner of each match-up is rewritten in the next round as the tournament progresses. Unfortunately, this representation requires a significant amount of space and makes it relatively difficult to get a quick overview of the tournament state since competitors take arbitrary paths through the static bracket. In this paper, we present AdaptiviTree, a novel visualization that adaptively deforms the representation of the tree and uses its shape to convey outcome information. AdaptiviTree not only provides a more compact and understandable representation, but also allows overlays that display predictions as well as other statistics. We describe results from a lab study we conducted to explore the efficacy of AdaptiviTree, as well as from a deployment of the system in a recent real-world sports tournament.