With advanced stats becoming more prevalent in hockey discourse, many terms that can fly over fans’ heads get thrown around. What do all the numbers, stats, and measurables mean these days? How does it impact the game we’ve watched for years?

Statistics in hockey help us understand the game through data. Goals, assists, and points are all examples of measurables that can be used to evaluate the effectiveness of players. As hockey evolved, so too have the ways to quantify players not just through counting stats. Like all stats, these new measures are useful in helping see the game from a different perspective. At the same time, it’s not possible to use them in isolation from the context.

For this series, we’ll be taking a look at some of the more commonly used advanced statistics, a small explainer of what it is, how it’s used, why it’s useful, and where it is limited. To kick us off is probably the most well-known advanced stat, Corsi.

What is Corsi?

The stat’s main objective is to track shot differential. Corsi takes shot attempts, missed shots, and blocked shots along with shots on goal into account both on offence and defence. It’s essentially a plus/minus tracker for shots that tries to communicate how often a player spends time in the offensive zone vs the defensive zone through the use of shots, thus indicating the puck possession of the team while the player is on ice. Generally speaking, shot attempts usually occur in the offensive zone, which means that any attempts for indicate your team has the puck, and any attempts against means your team does not have the puck. It can also be applied to a team overall over the course of a game or season.

Some of the most commonly seen Corsi values and their formulas are as follows:

Corsi For (CF) = Shots + Blocks + Misses
Corsi Against (CA) = Shots + Blocks + Misses from the opposing team
Corsi (C) = CF – CA

Corsi For % (CF%) = CF / (CF + CA); A positive CF% would imply that a player is controlling the puck more than he is not, while a negative CF% would communicate the exact opposite.

Corsi For % Relative (CF% Rel) = CF% – CF% of the team while the player is off the ice; If a player is making a positive impact, then their CF% Rel should be positive. If a player is making a negative impact, their CF% Rel should be negative.

Corsi Per 60 Minutes at Even Strength (C/60) = (CF – CA) * 60 / TOI

How Corsi is used

Corsi is primarily used as a proxy measure for puck possession. As mentioned previously, the fact that most shot attempts occur in the offensive zone means that the measures being tracked in this stat are related to the team/player having possession of the puck. In theory, a larger Corsi usually indicates a team took more shots than the other team, which in turn implies that they had more offensive pressure.

The stat could also be used as a greater sample size when looking at shots and the likelihood of scoring. Other metrics that we’ll be going over will be looking more directly at goals, but Corsi can be helpful when looking at goal events and the probability of them occurring. Because goals are relatively rare events over the course of a hockey game, using Corsi values as an expanded sample size allows us to view how many statistically significant Corsi events a player is a part of per goal scored. Through this, it allows us to have an idea of how effective a player is quicker than just looking at goals per shot.

How Corsi is useful

Individually, Corsi can help us identify players who are contributing positively at both ends of the ice. A CF% of around 50% usually means a player is making a neutral impact. This could be interpreted as a player being really good in the offensive zone and terrible in the defensive zone, or the exact opposite. A player who is below 50 CF% usually means that the player is making a net negative contribution to puck possession. A player who is over 50% can mean that a player is great in the offensive zone and doing enough in the defensive zone, or a player who is great defensively and doing enough offensively. CF% rel can help us identify the individual impact of a player on puck possession as well.

Generally speaking, a good Corsi has a positive correlation with points/60, with some of the best scorers in the league having good Corsi ratings. This makes sense, as greater offensive pressure and more shots should lead to more points, and having the puck is a big factor in that. What this correlation can also help with is predicting which players may regress in the near future. For instance, a player who is currently on a 5-game point streak with a low Corsi indicates that he is consistently getting out-pressured by the opposing team when he is on the ice. Thus, it wouldn’t be a surprise to see that point streak snapped because he isn’t in the offensive zone enough. The findings on Corsi have been that shot differential generally is a better predictor of future goal differential than past goal differential.

Team Corsi values are positively correlated with wins. Out of the top 5 CF% teams last season, only the Calgary Flames (2nd) did not make the playoffs, and expanding that to the top 10 CF% teams adds the Pittsburgh Penguins (9th). Conversely, none of the bottom 5 or bottom 10 CF% teams made the playoffs. The correlation does not mean causation, but the trends are ones to observe nonetheless.

How Corsi is limited

While Corsi is useful in contributing to the overall image of how effective a player is, the stat cannot be taken in isolation. For one, Corsi does not account for the quality of shots, treating all shot attempts as equal when clearly they are not. Corsi numbers also do not account for the quality of teammates. For instance, we saw Lane Pederson post some very solid CF% ratings alongside Elias Pettersson last season for a couple of games when it was clear that he was not the one doing the work there. This is when using other advanced stats can help in seeing the relative impact that he has, and what sort of contribution Pederson made when alongside the superstar Swede. On top of that, since the Canucks were generally a mediocre team, as a whole their players’ Corsi suffered in comparison to a team like Carolina. A team’s impact on a player’s Corsi is something that can’t be ignored or isolated, making it essential to look at relative player numbers with additional statistical categories. Corsi alone is not a good indication of how well a player is performing, but can be used in conjunction with other stats to get a better image of player performance.

There is an argument to be made that Corsi isn’t the best measure we could have to quantify puck possession. What would actually be better would be literally timing how often a player or team has the puck. For instance, a team trying to kill the clock would have plenty of puck possession, but not be registering any shots. This would show up on the Corsi charts as a team simply flatlining, doing absolutely nothing when in reality, they are dominating possession. With the development of player-tracking technologies, it would make sense that we could do this – except, nothing like that has been made available, at least in public statistical models. Until that happens, Corsi is the next best thing we have to equate puck possession as a stat.

Corsi works best when the game is at even strength and tied, which limits the versatility that this statistic has. A team will adjust how it plays whether it is trailing or leading, whether it’s on the man-advantage or on the penalty kill. The shifts in style can affect the Corsi values that a team generates, which may mislead people into thinking that a team is better than it actually is or vice versa. The predictive power of Corsi suffers as a result, meaning that more contextual data and information are needed in order to effectively use this statistic as a predictive tool.

This isn’t to say that Corsi does not have any value. Rather, it is one piece of the analytical puzzle that helps teams and fans understand the game better. The value that Corsi provides comes when it is used as a part of a whole, as further details in how hockey is broken down into data.