The Idiot's Guide to NHL Advanced Statistics, Part I

Jonathan Willis@jonathanwillisNHL National ColumnistDecember 10, 2013

ANAHEIM, CA - MARCH 21:  The scoreboard displays the statistics of Teemu Selanne #8 of the Anaheim Ducks after scoring his 600th career goal against the Colorado Avalanche in the second period during their NHL game at the Honda Center on March 21, 2010 in Anaheim, California.  (Photo by Victor Decolongon/Getty Images)
Victor Decolongon/Getty Images

Advanced statistics are not really all that advanced, and the ideas they convey are not especially complicated. Despite the idea that they require complex spreadsheets and lots of math, most of the concepts are based on operations like counting and simple division. 

Many, if not most, of the numbers actually reaffirm basic ideas that any non-numbers observer of the game will understand and agree with, even if he or she hasn’t considered them from a statistical angle. That can be helpful even for casual fans, since the numbers can serve as a sanity check for observation.

The significant roadblocks to accepting “advanced” statistics are not math or terminology or even fundamental disagreement between the eyeballs of a competent observer and the conclusions offered by the new statistics. The three most common roadblocks are threefold:

1. Fan rationalization

2. Over-reliance on traditional statistics

3. A lack of awareness of the effect of chance on hockey

EDMONTON, AB - NOVEMBER 15: Fans celebrate a goal by Nail Yakupov (not pictured) of the Edmonton Oilers against the San Jose Sharks during an NHL game at Rexall Place on November 15, 2013 in Edmonton, Alberta, Canada. (Photo by Derek Leung/Getty Images)
Derek Leung/Getty Images

There is nothing to be done about the first point; fans want to hear good things about their teams. This is a universal truth, and regardless of how factual a negative piece is (or how lacking in factual basis a positive piece is) some fans will opine loudly one way or the other based solely on their team colours.

There are few more reliable real-world examples of rationalization than fan message boards.

The second point is easily illustrated by an example: goals against average. GAA is the combination of two factors: save percentage and shots against per minute. Goalies have no control over the latter number, so basically GAA as a goalie statistic is a combination of save percentage and random noise. Save percentage isn’t just better; it renders GAA completely useless and has for more than three decades.

Yet GAA is quoted constantly. People cling to what they know. The NHL is rife with statistics both outdated and just plain useless (why hello, real-time statistics!) that are constantly repeated as if they had some actual value.

By far the biggest obstacle, though, is a reluctance to accept the role of chance in hockey. It’s the difference between what Nate Silver described as "The signal and the noise" in the title of his recent book: Every team’s record is a combination of true talent level and chance.

The best way to predict the future record is to, as much as possible, remove chance from the equation. 

This is what “advanced” statistics attempt to do. They provide unbiased fact, though of course there remains room for bias in interpretation.

They provide better information than the outmoded numbers that too many still rely on. And, most importantly, they try to reduce the noise and isolate the true talent levels of teams and players. 

That's why those numbers are important, not only for analysts but also for fans; they provide good information and can enhance understanding of the game. 

At The Team Level

“You are what your record says you are,” claims conventional wisdom. It’s not true.

Feb 21, 2013; Toronto, Ontario, CAN; Toronto Maple Leafs defenseman Michael Kostka (53) and defenseman Carl Gunnarsson (36) and center Tyler Bozak (42) and right wing Phil Kessel (81) celebrate a goal by forward James van Riemsdyk (21) on Buffalo Sabres g
John E. Sokolowski-USA TODAY Sports

Wins are relatively rare in the NHL (cue nodding in Buffalo and Edmonton). By that, we mean that expressing everything that happens in a hockey game with a single number means the loss of a lot of useful information.

A team can play well for 60 minutes and get a loss; a team can play badly for 60 minutes and get a win. Given enough time, it all balances out, but short stretches of a season aren't enough time and, in some cases, even a full year doesn't give a wholly accurate picture. 

More than that, wins aren’t a base statistic; rather, they are a function of goals for and against and, long term, a team with a good goal differential and a mediocre record is going to be better than a team with a mediocre goal differential and a good record.

For those readers confused by that sentence, I’d suggest applying for a job with the Florida Panthers right now.

Goals for and against are a better basis for conclusions than wins, but just like wins, goals are a product of something else. Goals for are generated by the combination of shots for and team shooting percentage; goals against are a combination of shots against and team save percentage.

Because it’s been shown—by yours truly for OilersNation.com, as well as by others elsewhere—that it’s hard for a team to take higher-quality shots than its opposition year in and year out, we’re left with two key drivers of goals (and consequently, wins): goaltending and shot metrics.

Goaltending is a highly individual statistic, which is why most analytics people focus on shot metrics to evaluate teams.

The key ones are:

Shots. Exactly what it sounds like.

Fenwick and/or unblocked shot attempts. Shots and missed shots.

Corsi and/or shot attempts. Shots, missed shots and blocked shots.

BUFFALO, NY - OCTOBER 21: Goaltender coach Jim Corsi of the Buffalo Sabres watches warmups before playing the Boston Bruins on October 21, 2008 at HSBC Arena in Buffalo, New York.  (Photo by Bill Wippert/NHLI via Getty Images)
Bill Wippert/Getty Images

The benefit of Corsi (named after Buffalo Sabres goalie coach Jim Corsi) is that one gets the largest sample possible, and the larger the sample the more accurate the results. Fenwick is a compromise; counting missed shots increases the sample size while at the same time not penalizing teams for good work getting into lanes in the defensive zone. 

The numbers are also modified in a number of ways.

Some prefer to look only at five-on-five situations, while others include four-on-four play. Because teams alter strategies in blowouts, others like to look only at situations where the score is tied or within one goal.

The numbers can also be expressed as a ratio (Team X had 52 percent of the Corsi events) or as a plus/minus (Team X went plus-26/minus-24).

All of these are details, included only to briefly explain the jargon; the truth is that it doesn’t really matter which of the shot statistics an analyst uses. Somebody using even-strength, score-tied Fenwick is very likely going to come to the same conclusions as five-on-five, score-close Corsi.

The numbers also, by and large, apply to special teams.

Consistently good power-play teams generate high shot volumes, while consistently good penalty-killing teams keep shots against low. Goaltending helps a lot in the latter case, and finishing ability makes a difference in the former, but the biggest factors are getting or preventing shots.

The important item is this: Good teams consistently win battles for territory and possession of the puck, and thus out-shoot their opponents while the game is still in doubt. Teams can win for a little while without doing that, thanks to a great goalie or a hot shooting run, but over the long haul winning teams are out-shooting teams.

That's why shot metrics—regardless of whether shots, Fenwick or Corsi numbers are used—should matter to fans.

It's easy to get carried away by a winning team, even when that team is doing things that will lead to losses down the road. It's at least as easy to get mired in a funk when a team loses a few, even when on the whole it's doing the right things.

Looking at shot numbers not only gives a fan a better idea of what's really happening, but it also softens the emotional roller coaster of an NHL season.

Those are the essentials at the team level. In Part II, we consider individual skaters.


The latest in the sports world, emailed daily.