Randy Sartin-USA TODAY Sports
The first post in our series examining the usage of advanced statistics in Hockey focuses on Corsi and Relative Corsi
I didn't write for most of the lockout, because I was pretty sure people didn't want to change their minds on what was happening. Let's be honest, most of us reached the point where we just wanted hockey back. I wanted my first post back to focus on the fact that I thought the Ducks would be better off pursuing a trade for P.K. Subban than seeking a trade for Ryan O'Reilly. To be fair, Subban and Cam Fowler on the point for the power play would have been the most beautiful thing in Anaheim since Paul Kariya to Teemu Selanne. But Montreal signed Subban and that idea went away.
Recent activity on the comments reminded me of a post I wanted to do earlier in the summer: Cam Fowler was getting a bad rap on his defensive play. I had a lot of statistical arguments lined up based on advanced stats. However, statistics are a funny thing.
In my academic life, I'm primarily opposed to quantitative studies in the social sciences. In fact, using science to refer to things like politics, social interactions, and communications is a little ridiculous to me. On the other hand, I feel that statistics are an important tool in understanding the performance of athletes.
Statistics are the primary way we examine performance and they have a comparatively low level of bias compared to personal observations. Stats are something we can all see and understand. They are indisputable in what they mean, so long as it is understood what they mean. Therein lies the difficulty and that is the partial purpose of this post.
Broadstreet Hockey did a great breakdown of some of the relevant stats and you can read that here. This post will at times sound similar because we will be describing the same statistics. You can catch another primer from the fine folks at Arctic Ice Hockey here. As I said earlier, this post will be slightly different in that I hope to incorporate the overarching qualitative vs. quantitative argument and I will be making an argument as to why Cam Fowler was one of the top defensemen in the league last season. However, that will all come in later posts in this series. It's a lot of information, so I'll be breaking it down into easier to consume segments.
What it is: Corsi is actually a pretty simple statistic. It measures the shots directed at the net while a player is on the ice versus the shots that are directed towards the opponent's net. It then expands them so that they are represented as a statistical value over a full 60 minutes. Whenever you read a corsi or rel corsi, it is being shown as a representation of the shot differential if a player played a full 60 minutes.
Relative Corsi compares a player to his teammates. When you see the number in the rel corsi column it is a representation of how many more shots were directed towards the opponent's net when the player is on the ice. If a player has a rel corsi of -12, that means that 12 more shots are directed towards his team's net per 60 minutes when he is on the ice as compared to when he is off it. In terms of calculation, it would be best to think of corsi as plus/minus but with shots instead of goals.
Limitations: There is still a luck factor in corsi because hockey is a team sport. A player can be in the right position and his teammates' mistake can lead to a shot on goal. It can be affected by how the coach uses a player, and it doesn't measure the quality of a shot which can be affected by good defensive play. Also, corsi can be affected by teams that trail, as they will throw everything at the net in hopes of tying.
Advantages: It limits the luck of plus/minus by dramatically increasing sample size. About 10% of shots in the NHL result in goals, so it's tough to say that goals are the sole representation of offense. Think of it in football terms: a defense that surrenders a lot of yards is inevitably going to give up a healthy amount of points. Corsi is the defensive equivalent of measuring yards instead of points. As a result, sample size is expanded and we have a statistic that is much more significant in terms of determining defensive effort.
What it means: Corsi is primarily used as a reflection of defensive proficiency, and it is a stronger representation of defensive capability than plus/minus. As a statistic, plus/minus is limited by the infrequency of occurrence and by a poor definition. If the defender mishandles the puck and it goes off the goalie's skate and in, that can hardly be called offense generated by the other team.
Corsi focuses on the truest representation of offense and quantifies it in a way that allows us to see where a player spends most of his time. A low or negative corsi means a player is spending a lot of time in his own end of the ice, and the opposition is generating shots in that time. A high and positive corsi indicates the player spends more time in the offensive zone and keeps the puck out of his end of the ice. There are other factors that can affect this such as Quality of Competition and Offensive start percentage, but they get their own post.
Finally, relative corsi is good for determining whether or not a player was highly susceptible to the effects of his teammates, and how important he is to his team's overall defensive proficiency. For example, a player can finish with a negative corsi on the season, meaning he was on the ice for more shots directed at his net than he helped direct toward the opposition, but still have a high relative corsi. What this means is that despite being on the ice for more shots against than for, his team was better off when he was on the ice. Furthermore, players who have positive corsi numbers but lower relative numbers might be benefiting from playing on better overall teams, and might not necessarily be pertinent to their team's defensive success.
Overall, corsi is one of the better tools in determining a players defensive proficiency.