The Puck Stops Here

I have written several sabermetrics and hockey posts this summer about adjusting +/- to try to cancel out team effects both as a rate stat and as a counting stat. I have used the rate stat method to adjust +/- ratings on the power play and the penalty kill. The important question to ask about this is how much can we learn from these numbers?

It is obvious by looking at the numbers (for example the even strength rankings) that this is not a ranked list of MVP candidates. The even strength adjusted +/- leader is David Perron of the St Louis Blues. There is no reasonable way to imagine that David Perron is the most valuable player on his team, let alone in the league. So what does David Perron's adjusted +/- mean? It means that he played very well in the role that he was given. The top players on these lists all were some of the best players in the league in the roles that they played. The bottom players all struggled and did poorly in their roles. This is one technique to identify such players. It is not the only technique. It is no truer that one can look at the point totals at the end of the season and assume it is a ranked list of MVP candidates (though the 2008 Hart Trophy nominees read as a ranked list of the top point scorers). It is possible to be a high scorer but detract from your team in defensive play and thus have low value. It is possible to be a low scorer play such a strong defensive game that you have a high value despite your lack of scoring.

The main problem with adjusting +/- ratings is figuring out how to normalize them. It is clear that +/- ratings are heavily influenced by the team in which a player plays. A player on a really good team will almost certainly have a higher +/- than a better player on a poor team. In order to attempt comparison between teams, it is necessary to adjust for that. There are several potential adjustments one can make. I think the most reasonable first step is to measure the difference that a player makes when he is on the ice against when he is not on the ice. This is a good starting point. It makes for rankings that are far better than the unadjusted +/- ratings. The problem is that players are compared to their team in the off ice comparison and not all teams are the same. Even for two players on the same team it isn't the same. Different players get used in different roles. They play with different players and against different players. This is the context under which the player's numbers must be looked at. These are things that can be quantified to some degree.

I think +/- is a very useful tool to learn about players. It is probably the starting point from which the most meaningful sabermetric hockey theory will be developed. This theory does not exist in any final form today - and likely will never reach the level that it exists at in baseball - but I think big developments will come in the future. Right now it is a useful statistic to help see the value of players. It is not the only tool. It is crazy to use these numbers as "final rankings". They are not. There is no one set of numbers that rank NHL player's values in any meaningful way. This is the idea behind the conversation in the power play +/- comments. Adjusted +/- is another tool to rate players. It is good to have as many tools as possible in your tool box. Sometimes one tool gets a job done that another tool does not.