clock menu more-arrow no yes mobile

Filed under:

In Defense of Corsi

Can analytics and the eye test method live in harmony among hockey fans? Here’s a little background on some of the more commonly used stats to try to bridge the gap between the traditional hockey fan and the numbers gurus.

New Jersey Devils v New York Rangers Photo by Bruce Bennett/Getty Images

There exists in the hockey world— and most sports really— a fundamental divide between those who advocate a more natural feel for the game and those who prefer a more hard copy, analytical view of performance and value. Maybe one day in a sports utopia we can all come together peacefully and agree that both have their use, but in the absence of Hockey Heaven I’d like to cover a bit of how and why advanced stats can be of value to even the most casual hockey fan, and why hockey writers often choose to reference them.

To be clear, I am not and will never argue that one way of interpreting hockey is better than the other. I think analytics provides an amazing way to view the outcome of a game or an individual performance in a concrete way that provides support to things we may see when watching the game itself. My objective in writing all this is that those who may want to better understand how and why we use the stats we do can hopefully gain a bit more insight from this, and those who may dismiss things that involve advanced stats may understand why we as hockey writers use them, and how much care is taken looking into the surrounding available data before choosing a single stat or two to represent a piece of the puzzle we are presenting.

First, a bit of a Hockey Stats 101 for those who maybe aren’t at all familiar in the stats we often use (Much of the data we talk about comes from Natural Stat Trick so I’ll try to base this off their definitions):

  • Corsi— one of the most popular metrics in hockey. Corsi is a measure of shot attempts that occur during a game. This includes shots that became goals, shots that were saved, shots that missed the net entirely, and shots that were blocked. Corsi for (CF) means shots taken by a team while Corsi against (CA) means shots taken against the given team. A Corsi For percentage (CF%) is a comparison of how many shot attempts the team had versus the other team. So a CF% of more than 50% means that team had more shot attempts than the other team.
  • Fenwick— Fenwick is all of the above shot attempts except those that were blocked by the defending team.
  • Scoring chances— this stat has some more advanced metrics behind it (which you can read a generally agreed upon background of here if you’d like) but basically, think of a scoring chance as a shot which has a relative chance to become a goal. Sites like Natural Stat Trick automatically exclude any shots taken from the neutral or defensive zone of the attacking team.
  • Scoring chance danger levels—obviously an 80mph slapshot from five feet in front of the net is more likely to go in than an 80mph slapshot from 50 feet away. As such, scoring chances are designated with a danger level that portrays how likely that shot was to become a goal. Shots from the slot and rush or rebound shots in front of the net can be considered high danger shots. A shot from in tight to the net, a blocked shot in the slot, or a rebound or rush from above the circles, at the boards or behind the goal line can be considered medium danger. A typical shot from the point, the boards, or behind the goal line would be considered a low danger scoring chance.

Now, all that’s great, but why do we care, how do we use them, and what size grain of salt should we take them with?

Most stats can not just be used as an indicator of the overall team’s performance, but also of individual players. On-ice stats show what shot attempts, scoring chances, and other metrics occurred while a given player was on the ice. Of course, all of this is subjective to other factors such as which zone they usually start in, what level of competition they usually face, or how good is the team they play on. For example, a players Corsi might be worse if they tend to start a shift on defensive zone face-offs because they are good defenders, whereas a poor forward might have good Corsi because they are only ever started in the offensive zone to counteract poor 2-way play. How do we know when this is the case? Well, because factors like these are also included in stats—this one is called offensive zone start percentage. It gives a good, quick to look at measure of how a player is utilized— if that number is around 40-50%, there’s probably not anything about his usage that could throw off their stats. If he starts in the offensive zone 85% of the time, his stats are most definitely being affected by that.

Similarly, a stat I like to look at for defensemen is overall scoring chances against versus high danger scoring chances against. A little bit of background, I am a defenseman, and have switched to pretty much exclusively defense for over ten years now. I’m a defense purist— I’m far less interested in how many goals a defenseman can put up if they can’t break up rushes, battle in front of the net, or make clean zone exits. Give me your Seth Joneses and your Dougie Hamiltons and maybe even an Andy Greene over an Erik Karlsson or Brent Burns style blueliner (I know, I know, one controversial take at a time). My point is, I don’t always like to evaluate defensemen by how many goals we get or shot attempts we generate when they’re on the ice—I have been there and I have lived the trauma of wingers who don’t know how to break out and leave us caught back in the defensive zone every single shift. I get it. I tend to use Corsi to compare defenders to other defenders on the same team because at the end of a game or season, you’re probably going to have a relatively even ratio of who plays with what forwards. For times that warrant a more in-depth look at a particular defender, I like to look at what proportion of the shots they gave up were actually high danger shots. A single defenseman can’t always control how much time the other team spends in their zone passing in circles and taking shots from the outside, but one thing you can often control on any given play however is where the player with the puck goes— a good defender forces the puck to the outside, whereas a poor defender gives up breakaways and fails to corral the forwards seeking rebounds in front of the goalie. Calculating the percentage of scoring chances a defender gives up that are high danger chances can be a way to evaluate their defensive zone play.

How do stats fit in with the eye test?

A lot of times I see fans argue that watching the game is better than reading stats, I see them reference things that can be and actually are quantified in stats, just as long as you know where to look and how to look at it. For example one might point out an extremely creative outlet pass that leads to an offensive zone rally that becomes a goal. The player who made the exit pass may not appear on the scoresheet, how and where can a number show you how difficult that pass was to find and make? How to you quantify ice vision? An important thing to remember about stats is that they become more valuable after a reasonable amount of time has passed (i.e. sample size). So you won’t see one good pass in one game on a data viz. What stats will show you is a pattern of finding and making good plays— those creative outlet passes will show up in zone exit percentiles, scoring chances for, even on-ice shot attempts for will be better for a player who is better at finding and making plays that turn into goals. The same goes for a player who routinely fails to find those plays.

For a good example, lets look at Nikita Gusev from this past season. Goose put up a respectable 13 goals and 31 assists in 66 games. Now obviously that shows he made an impact, but lets say I have concerns about his 200 foot game. Is his offensive drive costing us in mistakes that lead to breakaways for the other team? The magic 8 ball says no— Gusev’s shot attempts for while on the ice is better than the shot attempts allowed at a solid 54.94%. Most players on the Devils this season did not crack 50%—he is one of only 8 players on the team who played more than 10 games and has an over 50 CF%. In rates standardized by time on ice to shot attempts per 60 minutes, Gusev was rocking the highest shot attempts for on the team at 59— with only 60 against. This all tells us Gusev was not leaving the team high and dry in his search for offensive production.

Another, less obvious example is Jesper Bratt— he wasn’t the offensive cannonball that Gusev was but game in with a respectable 16 goals and 16 assists. He averaged less than 4 shot attempts more against than 4 in 60 minutes of ice time, meaning a CF% of about 48. Still good!

What would it look like if a player was a defensive liability to the team? Well, let’s take a look at the now non-Devil Mirco Mueller. Mueller allowed an average of almost 63 shot attempts against per 60 minutes with just 49 for, giving him a CF% of 43.81. Not good.

Another case is Miles Wood. Wood averaged 67.84 shot attempts against per 60 minutes, the highest of any regular lineup player. One could argue that Wood spent a significant amount of time on the ‘shut-down line’ with Blake Coleman and Travis Zajac, but Coleman actually had one of the lowest CA/60 on the team with a whopping ten less than Wood and Zajac had 6 less, so there’s some very strong fuel to the argument that Wood’s defensive play is just not good. His CF/60 is a bit lower than both Coleman and Zajac’s as well but not as drastically, which suggests his best usage would be primarily offensive zone starts and avoid playing him in situations like shut-down lines or when attempting to maintain a lead in the final minutes of a game.

My final word on this—

Advanced stats and data vizes are great ways to enhance our understanding of the game and demonstrate a player’s value and impact. They don’t take away from what we learn from watching the game. No one stat exists in isolation, and though a writer or sports commentator might present just one or two stats to support an argument, we know that it’s important to make sure there’s not other factors compounding the situation.

Your Take—

For those who know and like analytics, do you have any interesting or uncommon stats you like to look at when evaluating a player or game, or any good resources for stats you like to use? For those new to following analytics, is there anything I didn’t cover that you’d like to know more about? I’m looking forward to reading all of your thoughts, and thanks for reading mine!