CORSI IS DEAD (or is it?)
Note about the study
I, as with every person in the sport, have pre-conceived notions and biases about Corsi. However, I attempted to be as unbiased as possible and only present the facts for the first part of the study. I did not look at any data until I had already written the entire article minus the results and conclusion. This was my attempt to keep the study as neutral and objective as possible. My personal beliefs and opinions are reflected at the end of the study. Any appearance of an opinion throughout the rest of the study is purely accidental and unintentional. Please keep an open mind as you read this. You might be surprised at the results. I know I was.
Personal Existing Bias
I believe that Corsi is a statistic. Nothing more, nothing less. It helps to tell a piece of the story. It is not the story itself. The implementation of Corsi is flawed in the way I most often see it utilized. Using it as a standalone stat or with minimal backing ignores the incredible complexity of the sport of hockey. Many, not all, people that I see using Corsi would have me believe that it is the strongest indicator of success in the NHL and that there are only a couple outliers in the data.
I believe that the numbers and data will show a slightly different story. I don’t think that Corsi will be near as close to the end all, be all, that many make it seem. While it is foolish to believe that more shot attempts will not have an impact on overall success, I believe that the data will show that a team does not HAVE to have a strong Corsi For percentage in order to have success
Abstract
This study will attempt to identify whether or not a Corsi percentage above 50% will positively correlate with success, namely making the NHL playoffs, winning playoff games, and making it to the Conference Finals. While there are any number of ways to define success in the NHL, these three were the ones chosen due to easy ability to quantify success. Additionally, at the end of the day, winning is the ultimate form of success for a professional sports franchise.
For this study data from between the 2012-2013 and 2016-2017 seasons will be utilized. While this is a fairly arbitrary timeframe, it is a large enough sample size to eliminate isolated outlier years that could skew the data, and small enough that only the most modern game of hockey (shift towards speed and skill across four lines and away from having 3rd/4th line grinders) is studied.
Note: The 2012-2013 season was a lockout shortened season. Please take this into account when viewing data.
Introduction
Corsi is one of the most popular “advanced” statistics in the sport of hockey today. Since its inception as an official stat, Corsi has grown in popularity across the league. In the simplest of terms, Corsi is a measure of Shot Attempts, both for and against. While personnel across the hockey world fall onto a wide spectrum of viewpoints when it comes to Corsi, the Corsi debate GENERALLY falls into two camps.
The first side are people who believe that Corsi is the best or one of the best indicators of hockey success. Generally, they believe that a player’s success, talent, and contribution to a team can be accurately determined by their Corsi percentage. They also believe, that a team’s Corsi percentage is an accurate indicator of how successful a team is and will be. The other side of the Corsi debate believes that the implementation of Corsi is flawed. They believe that Corsi is not necessarily an accurate indicator of player or team success and that it must be viewed in relation to the many factors that determine the outcomes in professional hockey.
No matter where you fall on the spectrum of Corsi applicability, everyone would agree that more information is better. With more information you have the ability to make more informed and better decisions. However, in order to make that decision the information must be properly applied. That is where the two crowds split.
Brief History
Advanced Analytics have been around in one form or another for years. Most casual sports fans know about Billy Beane of the Oakland A’s, who was immortalized by Brad Pitt in Moneyball. In an attempt to compete with big market teams such as the Yankees and Red Sox, Billy Beane attempted to use mathematics to put together a team that could compete and win over the course of a long season. While he was somewhat successful in the short term, the long term impact was a statistical revolution that has reached out across all sports.
Hockey is no exception. Currently, the NHL has more statisticians and people with advanced analytics degrees than ever. Stats such as Corsi, Fenwick, Goals Saved Above Average, PDO, High Danger Scoring Chances, Zone start percentage, and Zone Entry with Possession Percentage are used more and more frequently to quantify a player’s skills and weaknesses.
For this particular study we are focusing on Corsi. Corsi’s creation as a statistic can be traced to Jim Corsi, the goaltending coach for the Buffalo Sabres and later the St Louis Blues. While there is some debate as to whether or not Corsi himself created the stat, or whether it was created by Chicago financial analyst, Tim Barnes, and just named for Corsi, there is no debate as to what the stat was made for. The stat was made and tracked in order to more accurately assess a goalies workload. While a goalie may only face 20 shots during a 60 minute game, that same goalie has to react to far more. Blocked shots, shots that miss the net, and actual shots on net are all events that require a goalie to move, expend energy, and wear down. This is what Corsi was originally tracked for. However, now Corsi has become a far wider reaching statistic that is used to help determine the effectiveness of individual players and teams.
Problem Statement
How does one scientifically determine whether or not Corsi is an accurate indicator of team success?
Everyone has their own opinion about Corsi, however, if there is no way to quantify or prove its applicability, the discussions just turn into infantile debates (looking at you Hockey Twitter). If we are to move on from these pointless debates, hard facts and serious study must be done. We cannot just point to a one game chart, a GIF about a not-top 10 play, or data without context. This study, while not comprehensive (no study is), will attempt to add a piece to the puzzle and further the discussion.
Hypothesis 1
A Corsi percentage above 50% will positively correlate with teams making the NHL playoffs during the period from the 2012-2013 season and the 2016-2017 season.
With this hypothesis I am predicting that teams with a Corsi percentage over 50% will have a positive correlation with making the playoffs. Between 12 and 16 out of the 16 playoff teams per year with a Corsi percentage over 50% will indicate a strong positive correlation. Between 8 and 12 out of the 16 playoff teams per year with a Corsi percentage over 50% will indicate a moderate positive correlation. Between 4 and 8 out of the 16 playoff teams per year with a Corsi percentage over 50% will indicate a moderate negative correlation. Between 0 and 4 out of the 16 playoff teams per year with a Corsi percentage over 50% will indicate a strong negative correlation.
Hypothesis 2
A Corsi percentage in the top 16 will positively correlate with teams making the NHL playoffs during the period from the 2012-2013 playoffs and the 2016-2017 playoffs.
Building off of Hypothesis 1, Hypothesis 2 predicts that there will be a positive correlation between teams in the top 16 of regular season Corsi percentage and those that make the NHL playoffs. A strong positive correlation will be indicated by 12 to 16 out of the 16 playoff teams per year being in the top 16 Corsi percentages. A moderate positive correlation will be indicated by 8 to 12 teams out of the 16 playoff teams being in the top 16 of Corsi percentages. A moderate negative correlation will be indicated by 4 to 8 teams out of the 16 teams being in the top 16 Corsi percentages. A strong negative correlation will be indicated by 0 to 4 teams out of the 16 teams being in the top 16 Corsi percentages.
Hypothesis 3
A Corsi percentage in the top 8 will positively correlate with Conference Finalists during the period from the 2012-2013 playoffs and the 2016-2017 playoffs.
Finally, Hypothesis 3 focuses on the “Elite” teams in the NHL. Teams that make the Conference Finals are, arguably, the top 4 teams in the NHL during any given year. With this in mind, Hypothesis 3 states that teams that make the Conference Finals will have a positive correlation with a very good (top 8) Corsi percentage. A strong positive correlation will be indicated by having 15 to 20 out of the 20 Conference Finalists having a Corsi percentage in the top 8 during that given year. A moderate positive correlation will be indicated by having 10 to 15 out of the 20 Conference Finalists having a Corsi percentage in the top 8 during that given year. A moderate negative correlation will be indicated by having 5 to 10 out of the 20 Conference Finalists having a Corsi percentage in the top 8 during that given year. A strong negative correlation will be indicated by having 0 to 5 out of the 20 Conference Finalist having a Corsi percentage in the top 8 during that given year.
Definition
Corsi can be measured in a number of different ways. The simplest way is to measure using Corsi For and Corsi Against. This is, simply, the number of shot attempts for and shot attempts against. You can also measure Corsi for and against per 60 minutes, Corsi for and against in 5v5 situations, and any number of other ways. While all of these different methods of Corsi measurement are helpful in different situations, for this study I will be using the Corsi For percentage in all situations. The way this percentage is calculated is by taking the total number of shot attempts for and dividing it by the total number of shot attempts, both for and against.
[youtube https://www.youtube.com/watch?v=sLpwYFBF29k]
For example: If the Rangers had 49 Shot attempts for and the Devils had 72 shot attempts against (sound familiar?) the Rangers Corsi For percentage would be 49/(49+72) or Corsi For/(Corsi For+Corsi Against). This comes out to a Corsi For percentage of 40.5%.
Methods
For this study I will be using data from between the 2012-2013 and 2016-2017 seasons. In order to prove or disprove the hypotheses I will only be looking at Corsi For percentages according to Naturalstattrick.com. While there are a number of great statistical sites out there including Corsica, Puckalytics, Hockey Reference and others, all Corsi For stats will come from naturalstattrick.com, unless otherwise mentioned, in order to maintain consistency and objectivity. This will not be overly complex as I am limiting the study to just one variable (Corsi percentage). Any additional variables would exponentially increase the length and complexity of the study. In addition, by isolating Corsi For percentage, I will be able to identify the good or bad of utilizing Corsi as a standalone stat.
Results
2012-2013 season
During the 2012-2013 season the Penguins, Canadiens, Capitals, Bruins, Maple Leafs, Rangers, Senators, Islanders, Blackhawks, Ducks, Canucks, Blues, Kings, Sharks, Red Wings, and Wild all made the Stanley Cup Playoffs. 12 out of the 16 teams were in the top 16 of Corsi For percentage while 10 out of the 16 teams had Corsi For percentages above 50%. In addition 3 out of the 4 Conference Finalists were in the Top 8. Out of the 16 playoff teams the Kings had the highest Corsi For percentage at 56.30% and were number 1 in the league. However, the number 2 team, the New Jersey Devils, did not make the playoffs at all. Finally, the Stanley Cup Final teams were ranked number 3 and 4 in Corsi For percentage.
2013-2014 season
During the 2013-2014 season the Kings, Blackhawks, Bruins, Sharks, Blues, Rangers, Red Wings, Lightning, Stars, Ducks, Flyers, Blue Jackets, Wild, Penguins, Avalanche, and Canadiens made the playoffs. Out of these teams, only 9 were in the Top 16 in Corsi For percentage and only 9 were above 50%. However, as in the prior season, 3 out of 4 of the Conference Finalists were in the top 8. The LA Kings had the highest Corsi For percentage at 56.82 and again the New Jersey Devils missed the playoffs while staying in the top 3 for Corsi For percentage.
2014-2015 season
During the 2014-2015 season the Blackhawks, Red Wings, Lightning, Predators, Penguins, Islanders, Jets, Blues, Capitals, Ducks, Wild, Senators, Canucks, Rangers, Canadiens, and Flames made the playoffs. 10/16 teams were in the top 16, while 12/16 teams had a Corsi For percentage greater than 50%. Additionally, 2 out of 4 Conference Finalist were in the top 8. As in the previous two years the number one team in Corsi For percentage was the LA Kings. However, this year, the Kings did not make the playoffs at all.
2015-2016 season
In 2015-2016 the Kings, Predators, Stars, Penguins, Ducks, Lightning, Blues, Red Wings, Sharks, Blackhawks, Capitals, Flyers, Islanders, Wild, Panthers, and Rangers made the playoffs. Out of these teams 12 out of 16 were in the top 16 and had a Corsi For percentage greater than 75% and 3 out of 4 of the Conference Finalists were in the top 8. The Kings, again, had the highest Corsi For percentage while getting bounced in the first round of the playoffs.
2016-2017 season
During the last season of the study the Bruins, Canadiens, Capitals, Predators, Sharks, Flames, Blackhawks, Maple Leafs, Blue Jackets, Blues, Penguins, Oilers, Ducks, Wild, Senators, and Rangers all made the playoffs. Out of the 16 teams, 11 were in the top 16 and 11 had a Corsi For percentage above 50%. However, in 2017 only 1 out of the 4 Conference Finalists had a Corsi For percentage in the top 8 or even in the top half of the league. As in all other seasons, the LA Kings had the highest Corsi For percentage with a 54.99 Corsi For percentage. As in the 2014-2015 season, the Kings failed to make the playoffs.
Overall
Out of the 80 teams that made the playoffs during the 5 years of the study, 54 were in the top 16 of their respective year’s Corsi For percentage. Likewise, 54 had a Corsi For percentage greater than 50%. Additionally, 12 out of the 20 Conference Finalists were in the top 8 of their respective year’s rankings. During each of the five years, except 2015-2016, at least one team ranked in the top 5 did not make the playoffs including two years in which the number one team did not make the playoffs.
Conclusions
Hypothesis 1
According to the parameters set out above and the data, there is a moderate positive while somewhat inconclusive correlation between having a Corsi For percentage above 50% and making the NHL playoffs.
Hypothesis 2
According to the parameters set out above and the data, there is a moderate positive while somewhat inconclusive correlation between having a top 16 team and making the NHL playoffs
Hypothesis 3
According to the parameters set out above and the date, there is a moderate positive while somewhat inconclusive correlation between having a top 8 team and making the Conference Finals.
Way Forward
As I have mentioned, hockey is an extremely complex sport and there are many variables that can effect the outcomes across individual games and across an entire season. This study has, hopefully, added a piece to the ever-evolving puzzle. In the future there are a number of ways to continue to getting after solving the puzzle that is success in the NHL. One such study would be to identify one, if any, singular statistic could be an accurate predictor of team success (I can already tell you that the answer is no statistic). Some of these potential statistics could include Fenwick (a similar, but better, (IMO) statistic to Corsi, Scoring Chances For and Against, High Danger Chances For and Against, or any combination thereof. Any study that is thought out, unbiased, and objective will help to continue taking the data that is already tracked and coming up with a more accurate story of how to be successful in the NHL.
My Personal belief
Hockey is an extremely complex sport and there are many moving pieces that affect the outcome of a game. Any, and all, statistics are helpful when determining effectiveness and success in hockey. More information and data is ALWAYS better. Corsi is one way to HELP measure pieces of the game. However, proper implementation and use of Corsi is key when utilizing the statistic. I personally find it silly when I see charts with single game Corsi numbers on them. While those charts can be used to back up an argument that a player had a bad game, it is like using a single game batting average in baseball. Yes, batting 1.000 is great for a single game but it does not tell the whole story. Did the guy rip the cover off the ball four times and hit two home runs or did he get a couple infield hits and a bloop single? Is he in the middle of a hot streak where he has batted .500 over 12 games? Or did he just get lucky in the middle of a cold streak. That is just in baseball, which is the easiest sport to use statistics to measure success. Hockey is far more complex.
It is this type of simple implementation of the stat that leads to misconceptions and perceptions that are not accurate. From 2013-2017 (2012-2013 season excluded for simple math) NHL teams averaged between 45 and 60 Shot attempts for (Corsi For). (Hockey-reference) Using basic math this means that the average player is on the ice for between 12 to 20 shot attempts per game depending on usage, zone start percentages, opponents, and a number of other factors. This means that just a couple shot attempts one way or the other can swing a player’s Corsi from dominant to dominated. For example, let’s say that Ryan McDonagh is on the ice for 20 shot attempts for and 15 shot attempts against. This gives McDonagh a Corsi For of 20, a Corsi Against of 15, and a very strong Corsi percentage of 57%. With a simple implementation this means that Ryan McDonagh is an elite defenseman.
Now, let’s go out on a limb here and say McDonagh is deployed to shut down Sidney Crosby or Alexander Ovechkin (I know, crazy, right?). Both Crosby and Ovechkin have a majority of their zone starts in the offensive zone (approximately 57% between 2012-2017). If McDonagh only plays when they are on the ice (just work with me) this “should” lead to approximately 20 shot attempt against and 17 shots for or a Corsi percentage of 45%. While not terrible, this is not a strong Corsi percentage, especially for a player who is, arguably, an elite defenseman. In fact this stat would seem to imply that Ryan McDonagh is a below average NHL defenseman.
Next, let us assume the same scenario, only this time McDonagh has a Corsi For percentage of 47.5% (19 shot attempts for, 20 shot attempts against). While at face value this would imply that McDonagh is still below average, he is, in fact, above average due to the zone start percentages.
Finally, let’s take this same exact scenario and apply some other filters. Ryan McDonagh, again, has a Corsi For percentage of 47.5% after having a defensive zone star percentage time of 55%. Now, lets add in the high danger chances for and against. When McDonagh was on the ice, there were 7 high danger chances for and only 5 high danger chances against. Now we are looking at a scenario where the Rangers created more high danger chances than the opposition (58%). In this scenario Ryan McDonagh, is finally shown to be the elite player that he is. Yes, he was on the ice for more total shot attempts, however only 25% of those were considered high danger chances against while 36% of them were considered high danger chances for like this one.
[youtube https://www.youtube.com/watch?v=LoFPXUem-h4]
When you put the whole scenario together and consider the matchups, zone start percentages, and type of chances (high danger vs non) you are able to see a much clearer picture.
Closing
In order to make the most of the Corsi statistic, it has to be properly used in conjunction with the rest of the relevant factors. For the past number of years the Rangers have not been a great Corsi team. In fact they consistently fall below 50% in the realm of 48%. However, the Rangers are consistently in the top of the league when it comes to high danger scoring chances (they lead the league this year with 347 as of 21 December). This could be an indicator that part of the Rangers strategy is to not settle for a high volume of shots but to work for higher quality shots that lead to more goals.
With the amount of data easily available to any person with an internet connection, it is simply irresponsible to take a single statistic and try to tell a story based on that statistic. There is so much information available. Take some time and put in some work and don’t fall into the trap of the easy narrative of “The Rangers suck because they consistently get out-Corsi’d (out shot-attempted).” I’ll close with what I started with. Corsi is a statistic. Nothing more, nothing less. It helps to tell a piece of the story. It is not the story itself.
More About:NYR Features