Introduction to Hockey Analytics Part 4.1: Possession Metrics (Corsi/Fenwick)

One of the things you sometimes hear from people talking about a hockey game is that one team is "controlling the pace of play." What this tends to mean is that one team has had control of the puck far more often than the other and has kept the puck in the opposing zone - and out of their own zone - for most of the game. When a team is able to do that, it seems like they are "outplaying" the other team, even if they've been yet unable to put the puck in the net. And the feeling seems to be is that if a team manages to control the pace of play - or control possession of the puck - then eventually the goals will come and they'll lead on the scoreboard as well in the spectator's minds.

Well this thought - the team that controls possession for most of the game is outplaying the other team and will most likely win the game in the end- actually seems to be as true as one would think. In other words, keeping possession of the puck - and more particularly, keeping it in the offensive zone - is indeed super important to winning in hockey.

Thus a good measure of how good teams are (and how good certain players are) would be to measure the amount of time the puck is in one team's possession, or perhaps the amount of time it was in the opponent's zone, per game. In Soccer (European Football), such a thing is tracked as an official statistic (See the "Time of Possession" measurement in this game here). Unfortunately, the NHL doesn't track such things.

So we're out of luck right? Fortunately the answer is no.

What Hockey Statisticians have done to compensate for the lack of a "time of possession" metric is to turn to rely on metrics based upon shot totals as a replacement. This should make intuitive sense - when a team has possession in the offensive zone it aims shots at the net, while when it doesn't shots go toward its own net (and if a team has possession in its own zone, shots go nowhere). So while we can't tell exactly how many minutes each team controlled the puck or the minutes that the puck was in the opposing zone, we can make a good guesstimate from shot totals.

Enter Corsi: Corsi is a statistic that simply the equivalent of +/-, but using shots instead of goals. Note that I used the word "shots" instead of "shots on goal" - Corsi doesn't care at all whether the shot was "on goal," whether it was blocked, or whether it missed by 5 feet. This is because Corsi only cares about shots because shots show what zone the puck is in - and missed and blocked shots do that just as well as shots on goal.

Now, Corsi statistics don't say anything specific about how a particular player plays: rather, they simply say how the team performs over a given time span by giving an approximation of how much the puck has been in a team's offensive zone as opposed to its defensive zone. When applied to players, just like +/-, Corsi can be used essentially to measure the value of ALL of a player's contributions to his team while the player is on the ice*.

*Corsi metrics, and all possession based metrics, can't measure the impact a player has on the ice when that player is actually off the ice. In other words, if a player's game involves making the opponent more tentative through hard hits or to make the opponents less likely to hit star players for fear of retaliation, that type of contribution cannot be measured by corsi.

Well sort of. See Corsi is a good deal affected by context, as is every other hockey stat (See Parts 2.1-2.4 of this primer). And since Corsi is measuring team performance, you have to take into account Teammate Quality before you can easily use it on a specific player. In addition, two other factors are major contextual factors that need to be taken into account when you use corsi.

The first is where players start their shifts - in other words, their Zone-Starts %. If players are sent out more often on faceoffs in the offensive zone than defensive zone, their corsi numbers are likely to be higher than if it was the other way around (We've talked about this before). Thus defensive specialists often can have quite negative Corsi (see Manny Malhotra for an example) simply because they start their shifts much more often than not in the defensive zone. There is a way to compensate for that and normalize corsi to account for the impact of a player's zone-start %, but no hockey site has such normalized corsi values readily available.*

*If you want to normalize the corsi values by hand, you can do that. Each additional defensive faceoff taken while a player is on the ice lowers his total corsi value by 0.8. So to compare players with different zone-starts, you can simply "add" offensive faceoffs to one player till the two players have equivalent zone-starts, adding .8 to the total corsi of that player for each faceoff.

The second factor is the score and time remaining in the game. These factors are often called Score Effects. Essentially, when a team is up late in the game, it tends to play more defensively (sometimes going into a large defensive shell), and stops taking shots in favor of simply trying to hold on to the lead. I'll probably go over these factors in a later post. But for now, just understand that to use corsi most effectively, you're best off using corsi numbers from when the score is tied or for when the score is close and not too late.

These are not the only factors that you need to account for, but these are the biggest to keep in mind, and both of these factors can be accounted for.


Now, there if you go to places like, you'll find there are a lot of different statistics with the word "Corsi" in them. What do they all mean? Which should you use? And that'll be the subject of the next post in this series, Part 4.2: an explanation of the various possession metrics out there.


The Intro to Hockey Analytics/Advanced-Hockey-Statistics Primer so far:

Part 1: - What is the field of Hockey Analytics and Why Might You be Interested?
Part 2.1: - The Importance of Context Part 1 - Time on Ice
Part 2.2: - The Importance of Context Part 2 - Evaluating the Difficulty of Certain TOI through QUALCOMP and Zone-Starts
Part 2.3: - The Importance of Context Part 3 - Evaluating (and Compensating for) the Effect of Teammates via QUALTEAM and Relative Measures
Part 2.4: - The Importance of Context Part 4: The Concept of the Replacement Level Player
Part 3 - The Perils of Sample Size

<em>Submitted FanPosts do not necessarily reflect the views of this blog or SB Nation. If you're reading this statement, you pass the fine print legalese test. Four stars for you.</em>