Animal play is a topic that has received a good detailed attention in recent years (see and and, for example) because coming to an understanding of this behavior is central to learning about the evolution and development of play and other behavior patterns and also neurobiological correlates. In addition, learning about what animals actually do when they play is central to furthering our understanding of the evolution and development of moral behavior (fairness and cooperation) in animals (see, for example, Wild Justice: The Moral Life of Animals, "Moral in Tooth and Claw", and "The Ethical Dog").
Various theories have been offered about why animals play, and there's no one explanation that fits all examples of animal play. Play is important in social development, physical development, and cognitive development and also may be training for the unexpected (Spinka, Newberry, and Bekoff 2001). Based on an extensive review of available literature, my colleagues Marek Spinka, Ruth Newberry, and I proposed that play functions to increase the versatility of movements and the ability to recover from sudden shocks such as the loss of balance and falling over, and to enhance the ability of animals to cope emotionally with unexpected stressful situations. To obtain this "training for the unexpected" we suggested that animals actively seek and create unexpected situations in play and actively put themselves into disadvantageous positions and situations.
Animals may also play because it's fun and feels good, during which time they also are benefiting from the activity itself. This video of a Siberian husky playing by himself (or herself) tells it all, and no other explanation seems plausible. It's a dataset in and of itself and offers valuable insights into why animals play.
The teaser image can be seen here.