There are two main reasons why the adoption of analytics in hockey has been slow compared to other sports. First, the game of hockey has more of a flow aspect compared to other sports. Therefore, compared to other sports, more complex statistical methods are required to accurately describe the game of hockey with data. Second, a deep understanding of the game is required to create analytical models of sporting contests. Compared to other sports, there are fewer people with the sufficient skill set to create the complex analytical models required to describe the game of hockey with data.
Therefore, because the game of hockey is complex, conventional types of analyses do not yield accurate results; and because managers are very good at evaluating talent, they have not adopted analytics derived from these conventional types of analyses. On the other hand, complex statistical methods require a specialized skill set to complete. In addition, sometimes complex analytical models are more difficult to describe than traditional types of analyses. Therefore, managers have been forced to decide between conventional types of analyses that do not accurately describe the game of hockey and more complex analyses that are difficult to understand.
Fortunately, although we use complex statistical analyses to derive our results, we have developed a method for describing our models to managers. Overall, our model for evaluating players consists of the following three stages: Predicted Goals Scored™(PGS™), Predicted Wins™(PW™), and Contribution to Winning™(CW™). Each stage uses increasing levels of statistical analyses. These stages, in combination, describe the game of hockey with data well. While each level of the model uses complex statistical methods to derive the results, in isolation, understanding the results is intuitive.