When I talk about ways of evaluating players, I often add that metrics for players can be combined in meaningful ways for teams. Here are two quick examples. First, ratings from different models can be combined to show who’s most important to a team. The graph below shows 2014-15 ratings from two of my models (explained here) along with Shapley values for the club I support, Newcastle United. Shapley values aren’t as persistent from season to season as the other models’ ratings, but they can help to suggest who’s a misfit on a team and who’s pivotal. The diagonal lines…
Read more →