Using Corsi to measure goals for the Rangers

AP Photo/Gene J. Puskar

If you’ve been reading us for a while, then you know we are proponents of the new stats that have popped up over the past few years. One that has really taken hold in the community, and led to the hiring of several “stats guys” around the league, is Corsi. Corsi is a measure of shot attempts (on net, missed, or blocked) taken for or against while a player is on the ice. It’s basically the plus-minus of these shot attempts. If Player X is on the ice for three shot attempts for and one shot attempt against, then that shift he was +2 for his Corsi (also a 75% Corsi). Fairly simple concept.

Ignoring some of the noise associated with these stats, the biggest complaint that comes up with the possession metrics is that it doesn’t measure goals, and goals win games. It’s tough to argue with the logic, because it’s partially true. Great work has been done proving that possession directly correlates to wins with large enough sample sizes, but it still doesn’t measure goals. Pucks in the net are what matter.

Luckily, the guys at Hockey Graphs have begun breaking ground on this front. They did a lot of the legwork in measuring goals via possession numbers, and that post is definitely worth a read in its entirety. But to summarize: Players can be separated into four buckets (forwards) or three buckets (defense) based on ice time. Each player falls into a bucket based on ice time, and you can use their Corsi% to project a goal differential across the season.

We can use the work done here to look at the expected goal difference for each player on the Rangers from last season (all numbers at even strength):

Forwards

Defense

First thing to notice here is that based on possession numbers, there were only two players that were expected to finished with a negative goal differential: Kevin Klein and Dominic Moore. Now, that’s not to say that they were ineffective in their roles, because while this is certainly great information, it –at the moment– doesn’t take into account other factors. The article says that zone starts, quality of competition, streak effects, etc. That’s a long-winded way of saying that Corsi on its own isn’t the best way of using these stats.

This isn’t a predictive measure by any means, something the author also states. The value here is using the raw puck possession numbers to evaluate how players performed, while making note of the role they played. For example, Dom Moore is expected to have a negative expected goal differential because the only regular to start more shifts in the defensive zone was Henrik Lundqvist (full zone start information here). You can combine what you see above with the zone starts to see how effective a player was last season.