Several years ago we welcomed, some more enthusiastically than others, a new era in sports analysis stimulated by the success of Moneyball. Published in 2003, the book Moneyball: The Art of Winning an Unfair Game chronicled the case of the Oakland Athletics, who relied on a new type of analytical approach to building a competitive team despite Oakland’s disadvantaged revenue position.
Hockey followed suit creating new metrics in the quest for Moneypuck. Most of the measures were just another way to quantify hockey. However, this sport is much more complicated than can simply be explained by more measurement.
Providing additional ways to quantify a player does not necessarily lead to more accurate evaluation. His actions have to be defined as well. To me, those measurements are like the addition of color to the black and white TV.
For many years, the data available was rather simplistic as were the conclusions: a scorer recorded a lot goals, a playmaker tallied a lot of assists and a physical player had many hits.
And we were introduced to Corsi.
As we know, Corsi is the measure of shot attempts for (shots on goal, blocked shots, missed shots) at even strength less shot attempts against. The abbreviation adopted by the NHL is SAT.
The proponents of this metric began to evaluate all the players using it, believing that they had found the holy grail, the magic number of the modern era. The value of players with a low Corsi was dismissed.
Corsi has become synonymous with possession, but to be clear, I present a quote from our Editor-in-Chief: “Corsi only infers possession, it does not measure it.” — Rick Stephens
It is taken for granted that if a team tends to shoot more towards the opponent’s net than the shot attempts that they allow, they are most likely possessing the puck more often. It is an assumption that can be correct and with which I can partially agree.
But where it goes a little off the rails is taking the assumptions several steps further. The advocates of the measure state that more shots attempts yield more goals and more wins. That’s a leap too far and this is where I make my departure.
Let me present a table from last season. Corsi For Percentage (CF%) is calculated using this formula: Corsi For / (Corsi For + Corsi Against). Teams highlighted in yellow qualified for the playoffs.
As you can see, a team can dominate Corsi, (Kings, Hurricanes, Flyers, Panthers) but not be successful during the regular season or the playoffs. To further the point, the Cup-winning Pittsburgh Penguins are well down the ranking in terms of CF%. It is clear that this measure has little predictive value of success.
On average the rank differential was 7.5 on average for 2016-17. It can be wildly inaccurate with no correlation.
Why is that so? Because this statistic simply compiles shot attempts without defining their quality.
Let’s take a look at two examples using these fictional play-by-play calls.
- “Tyler Seguin enters the zone against Shea Weber on the side boards. Seguin tries to sneak past Weber, but is choked off. From the bottom of the faceoff circle, Seguin takes a wrist shot from a sharp angle directly into the torso of Carey Price. The Vezina-winning goaltender easily makes the save.”
Statistical result: Weber and his Canadiens on-ice teammates receive a shot against on the Corsi chart, despite the fact that Weber positioned himself perfectly to force a poor-quality shot eliminating a scoring chance. Seguin and his Dallas mates receive a shot for.
- “Alex Galchenyuk enters the zone, dangles Ryan McDonagh, and makes an amazing saucer pass to Pacioretty waiting in the slot. Pacioretty one-times the pass but is foiled by Henrik Lundqvist who makes a tremendous glove save using his exceptional reach and lateral movement.”
Statistical result: The Canadiens players on-ice receive a shot for on the Corsi chart while the Rangers are charged with a shot against.
It is clear that these two plays are completely different but not in the eyes of Corsi. And we could make the case even more strongly by including an example where the puck did not even reach the net.
There are so many aspects of the game of hockey that Corsi ignores: the quality of the shooter, the angle of the shot, the distance, the precision, the speed of the action, to name just a few. These are aspects that everyone who plays the game and follows the game more than casually know innately.
It is obvious that Weber should be credited for his good coverage allowing an easy shot for his goaltender in the first example. Galchenyuk should be recognized for his ability to make a very difficult pass setting up a high danger scoring chance against a good opponent.
With it’s inability to distinguish between these two very different plays, how can we trust Corsi to describe the effectiveness of a team or the impact of a player? As mentioned, it is not even reliable as a measure of possession, just an approximation at best. And that utility will become obsolete the day that the NHL begins using tracking chips in jerseys.
An interesting calculation is the point sharing system presented by Hockey Reference. It is derived from the Win Shares system in baseball. This is an estimate of the total standing points contributed by the player to his team.
The total of each players point shares is close enough to the final standing. Additionally, it shows the importance of a player has for his team. On the other hand, the measure doesn’t really explain how each player contributes to the game.
Let’s take a look at the top ten players in terms of Point Share (PS) for the 2016-17 Canadiens lineup:
Is anyone surprised to see Carey Price on top? That said, was Nathan Beaulieu really such a cog to the success of the Habs last season?
At this moment, the magic bullet hasn’t been discovered. Nevertheless, there are plenty of advanced stats we can have fun looking at: the pass/shot in the danger zone, the puck entries in the offensive zone, the carry-in against or the breakout passes success rate.
There are so many dimensions to this game of hockey that may not present themselves in other sports and are not easily captured by simplistic measures. Each has to be taken with grain of salt and certainly not in isolation.
We can all acknowledge that there is still a lot of work to be done in this area. And that it is silly to dismiss an experienced eye over the long term as part of any evaluation. In this series we will look at other tools.
But for now, tell me how you follow hockey. Is there a measurement that is meaningful to you or do you prefer as of right now, do you have some favorite advanced stats or do you trust your own judgement?