A Beginner’s Guide to Advanced Hockey Analytics

I know, I know. It feels like it happened so fast, you barely noticed it and you feel like you got left behind. I get it.

But hockey is a stats game now. Sure, there are some holdouts in certain mostly western Canadian media markets, but teams are increasingly being built using spreadsheets and giant dweebs who never Played The Game.

To be sure, nobody is advocating completely refusing to actually watch the games, but rather to use analytics to see what cannot be remembered. People love the eye test, and I get why. But our brains have limits. When there are 13 moving things on the ice at any given time, tracking one or more of them and developing a perfect understanding of their tendencies and habits, especially in the greater context of the on-ice situation is absolutely beyond our comprehension. To paraphrase Moneyball, the difference between a .250 hitter (probably starter quality) and .300 (All-Star) is one hit per week. Your brain will not remember that hit. You need to read the stats to know if the player has it.

Hockey stats for the most part are in their infancy, roughly equivalent to a batting average. You’d think this wouldn’t exactly be controversial but here we are.

There are a few terms we’re going to break down for you here and it’ll be a rough understanding of a very basic group of stats, but it’ll get you started.

First is Corsi. Corsi is meant to measure 5 on 5 possession time. Why 5 on 5? Because most of the game is played 5 on 5.

Hockey is an exceedingly fluid game. This is what makes quantifying it so difficult. Questions which are extremely easy to answer in other sports, such as “who has the ball/puck?” are actually fairly difficult to answer for large portions of hockey games. Whereas in football or basketball they are very straightforward.

Corsi is a straight measurement of shot attempts. Whether the shot is on goal, off the post, blocked, off the net, deflected, or any other myriad of things that could happen to it.

Now I know what you’re thinking, dear reader!

“You can’t just sling the puck at the net all the time and expect it to go in. You need to create genuine chances!”

To this end, you are entirely correct. Corsi is a stat that must exist in the periphery of the game. Players must play without regard to it. And they do. They play to score goals and defend their net. If players started trying to actually game their possession numbers, their game would suffer and they would be less effective. Because of the aforementioned fluidity of the game, Corsi is nothing more than an attempt to quantify the question “Who has the puck?” That’s it. That’s what the stat is for.

EXAMPLE: Team 1, let’s call them the Lightning are playing a game against Team 2, we’ll call them the Leafs.

In the first period the Lightning have 30 shot attempts. 15 on goal, 10 missed the net, 5 blocked.

The Leafs have 10 shot attempts, 3 on goal, 5 missed the net, 2 blocked.

The raw Corsi would be expressed as 30-10. But no one uses that because Corsi is meant to provide a bigger sample size than, say, wins or goals. As you can imagine, these numbers get large very quickly. Therefore, they are typically expressed as a percentage. So in this case, the Lightning had “75% possession.” A good team is above 50. Simply meaning they have the puck more than their opponents. A great team is around 55%. As you can see, the margins are very thin.

Corsi ratings are also applied to players, the same as they are applied to teams. So let’s reimagine our example for a player analysis.

Say in our previous example that Auston Matthews on the Leafs is on the ice for 5 of the shot attempts for his own team, and 5 of the shot attempts against. Auston Matthews would be credited with a 50% personal Corsi. That is a strong deviation from his team’s lackluster performance. If that’s the case, it’s clear to see that Matthews is likely not the problem. When he’s on the ice, his team is three times as effective as it is without him.

Most possession stats are different mutations on Corsi. Some measure shot attempts from “high danger” areas. This is typically referred to as the “home plate.” The home plate goes from the left and right edges of the crease, diagonally out to the faceoff dots, and straight up to around the tops of the circles. (There are slight variations on this, however.) Why are those the “high danger areas?” Because that’s where a vast majority of goals are scored. This ver

The last thing we’re going to discuss is PDO. What does PDO stand for? I don’t know. PDO is an attempt to quantify “luck,” which plays a tremendous role in the game of hockey. It’s fairly simple. You simply add a teams 5 on 5 shooting percentage to their 5 on 5 save percentage. Wherever that number lands relative to 100 will somewhat reflect the “luck” of the team. Sometimes a team hits a streak of luck and the puck just keeps going in. Not because they’re generating more possession or chances, they’re just getting lucky. PDO is meant to detect whether a streak of winning or losing is based on quality of play, or just “running into hot goaltenders.”

There are obvious shortcomings with PDO to be sure. Particularly that having an above-average goalie is going to make a team’s save percentage above average. And having a team with higher than average shooting talent will improve a team’s overall shooting percentage. There is still a general range where being a really good team can land you. That number is around 101.5. Any higher and you’re getting some unsustainable luck.

For example: At time of writing the Tampa Bay Lightning lead the league in PDO at 102.66. This is the sum of their team save percentage at 93.66, which is more regularly expressed as .937. They are also second in the league in team shooting percentage at 8.9%. That save percentage is absolutely ludicrous, and far away from the league median closer to .925. But the eye test will tell you that Andrei Vasilevskiy is easily one of the best goalies in the league, if not the very best. The shooting percentage is also well above the league median 7.5%. That is expected given their offensive talent in Kucherov, Hedman, Stamkos, and Brayden Point. But it remains outside the norm. One would expect these numbers to regress slightly.

To complete our analysis, we’ll compare the Corsi and PDO of the Lightning.

Currently, the Lightning possess the top PDO, but ALSO are 8th in the league with a 51.94%. So the simple analysis is that they are holding the puck relatively well, but are likely getting some luck in both nets. If you are a newcomer to the advanced stats community, this is the analysis you’re going to have, and wrangle with. Is the PDO because of superior talent, luck, or a combination of both? The eye test says it’s likely a bit of both. They have arguably the best forward pair in the league in Stamkos and Kucherov, along with possibly the best goalie in the league in Vasilevskiy.

Conversely, Colorado boasts a Corsi of 48.32%, but a PDO of 101.68. The eye test says Nathan Mackinnon is having one of the best seasons of all NHL players, but that PDO is backed by an 8.9% shooting percentage. Is that team full of genuine talent to make that shooting percentage sensible? The eye test says no. Colorado may be a team that is relying on luck to fuel their incredible run.

I should reiterate, these stats are FAR from the be all, end all of hockey analysis, but these are factors that a relative newcomer to the advanced stats community should be able to learn and lean on as they try and guess who are contenders and who are pretenders.


Leave your HOT TAKES here:

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s