UGG Baratas online. Australia UGGs Outlet España.

Your #1 Source for Nashville Predators News, Updates, and Opinions – A Nashville Predators Blog

An Introductory Guide to Advanced Hockey Analytics


Oct 14, 2013


For the last seven-or-so years, there’s been a movement to see beyond the familiar stats so readily available for time immemorial. While goals, assists, points, and penalty minutes are useful stats, they don’t tell you much about context. Plus/minus has long been decried as a borderline useless measurement. There are too many variables to taint those numbers. For example: a player makes a poorly-timed line change, resulting in a goal. The player replacing him on the ice, well out of the play, receives a minus and the original player’s stats remain unblemished. So out of that inadequacy, a small group of fans, bloggers, and the odd statistician developed a more accurate means of measuring player and team efficiency.

I don’t want to re-invent the wheel or re-hash the history of advanced statistics; several people have written some great, in-depth articles to that end (see the “Recommended Reading” section at the end of this article). My hope is to provide a simplistic, Predators-centered explanation of some of the terms that you’ve seen thrown around at a conceptual level. You may find that if you can grasp a few general ideas, you’ll be able to make quick sense of these measurements, and maybe gain a better perspective of the game you just watched. This guide is meant to be a companion to a new series that we intend to embark upon here at The Predatorial, in which we’ll give a “How They Did” breakdown after each game, from an advanced-stats perspective.

At its core, advanced hockey statistics can be distilled into a few simple tenets. “Corsi,” “Corsi-Relative,” “Fenwick,” “Fenwick-Close” are just a means to measure the one, tried-and-true holy grail of predictable success: puck possession. History shows us that teams that possess the puck tend to win more than they lose. If your team plays most of the game with the puck and controls possession, the odds are good that on most nights they’re going to win. If they can maintain strong possession numbers over the course of a season, their chances of a deep playoff run are that much better. Teams that are puck-possession beasts, year in and year out? The Detroit Red Wings. The Los Angeles Kings. The Chicago Blackhawks. The Boston Bruins. Notice a pattern? Conversely, poor possession teams tend to fare…well, poorly. That’s not to say there aren’t statistical anomalies.

Can you be a bad possession team and still make the playoffs? Sure..the Minnesota Wild did it last season and were swiftly handled by the Blackhawks. Another example of a historically weak possession team that has overcome the odds to make the playoffs? The Nashville Predators. Taking on another layer of advanced statistics will show you that there are even ways to measure “luck.” Unlike puck possession–the seemingly ironclad predictor of effectiveness of a team–luck has a way of running out eventually.

To begin to understand the measurements of possession, the foundation building block is “Corsi,” named for Buffalo Sabres goaltending coach Jim Corsi. Corsi was the first to opine that shots of any kind, even blocked shots and missed shots, not traditionally events that increment the shot counter, added to the workload of a goaltender. It makes sense, a goaltender has to react in those instances. It would be like omitting wild pitches and balls from pitch count, and only counting strikes. The forefathers of the advanced statistics philosophy posited that if this new inclusive measurement of shots represented a more accurate metric of the strain on a goaltender, the players generating this workload could also be measured by it. Essentially, if you read no further, Corsi alone will put you about 75% ahead of the game when it comes to advanced statistics. With that simple concept, you can tell a lot about what actually happened en route to the final score.

The “Corsi rating” is calculated as a basic plus-minus of all shot attempts, called Corsi events. If Patric Hornqvist (historically a very good Corsi player on a weak Corsi team) is on the ice for 12 Corsi events for, and eight Corsi events against, he’s a +4 player.

Most often, Corsi rating is expressed as a percentage. Of the 20 Corsi events generated by the Predators and their opponent while Hornqvist was on the ice, his team accounted for 12, or 60%.

Similarly, you can look at the Corsi rating for an entire team. This is a good means to see who controlled most of the play. To simplify: the team that’s generating most of the shot attempts is likely possessing the puck, and thus controlling play. Over the course of a game, say the Predators generate 60 Corsi-For events, while their opponents only generate 30 Corsi-Against events. The Predators handily owned possession, 60/(60+30) = ~67%. Does that mean they won the game? Not necessarily, there’s no way to remove “puck luck” from the equation totally, but there is enough statistical evidence to definitively say that in most cases, teams that ride a positive Corsi rating are going to achieve success more often than not.

While Corsi is the most popular of the advanced metrics, you may also see “Fenwick” used, which is the same concept as Corsi, but removing blocked shots from the equation. In other words: a goal, a shot, and a missed shot all constitute a “shot attempt.” Since things like power plays and penalty kills obviously skew Corsi/Fenwick, we tend to use “5 on 5 play” as our sample. In addition to standard Corsi, there’s also an interesting statistic called “Corsi Relative.” As the name would suggest, this is the team’s Corsi relative to that player, measured against the time the player is not on the ice, extrapolated over 60 minutes.

For example, if Patric Hornqvist has a +16 Corsi-Relative rating, that says that the team generates 16 more Corsi events when Hornqvist is on the ice than when he’s not on the ice. On the other hand, if Filip Forsberg has a -12 Corsi-Relative, the team is likely to give up 12 more Corsi events per 60 minutes when Forsberg is on the ice versus when he’s not. Like standard Corsi, positive numbers are good, and the higher the better.

With that easy-to-understand concept, you could stop reading here and carry the salient building block of puck possession with you. But if you want to take it a step further, you may ask yourself, “aren’t there extenuating factors that affect a player’s Corsi rating?” There definitely are, but fortunately we can account for those. If you cross-reference a player’s Corsi rating with a couple of other statistics, you can begin to make sense of why a player’s rating might be higher or lower. Two statistics in particular are useful in this regard: deployment (zone starts) and quality of competition (QOC or QualComp).

For example, Paul Gaustad is a poor Corsi player over his career, but think about the situations in which he’s most often used: taking faceoffs in the defensive zone against the opponents’ scoring line. Since he is used in defensive situations most of the time, it stands to reason that he’s going to be on the ice for more Corsi events-against than events-for. So, in this instance, Corsi alone doesn’t tell us the whole story. We need context, and for that, we can look at his deployment percentages. If Gaustad has 12 total zone starts over the course of a game, perhaps we see that eight of those were in the defensive zone, two were in the neutral zone, and two were in the offensive zone. At that point, his Corsi rating becomes a little easier to explain.

To take it a step further, we can look at a couple of measurements to see what type of players he was up against, quantitatively a measurement called “quality of competition.” This is typically a representation of the average Corsi rating or Corsi-relative rating of the players the opponent iced against a particular player. If Mike Fisher’s competition has an average Corsi rating of 60% and an average Corsi-Relative of 12.8, you can see that he’s going up against good players. Similarly, at first glance, a player carrying a 75% Corsi rating may look great out of context, but may be tempered by a relatively weak Quality of Competition and low defensive zone starts.

Now that you’ve mastered the metric of puck possession, you may wonder about “exceptions” and anomalous results. As I alluded earlier, sometimes good old-fashioned “bounces” will go a team’s way. Post-2007, the Predators have historically put up below-average possession numbers, yet they still made the playoffs on the back of a strong regular season. Many pundits didn’t take them very seriously, and in a lot of cases it’s because they were aware of the underlying advanced statistics, which suggested there were a few unsustainable factors at play. Barry Trotz, in years-passed, would point out “shot quality” when defending players like Sergei Kostitsyn and Martin Erat. Certainly, their shooting percentages supported his statements. However, with everything we’ve covered to this point, you should understand that shot quantity trumps shot quality every single time.

To this end, we have one more “fancy” stat to discuss: PDO, which essentially measures a player’s luck. The way it’s calculated is simply the team’s shooting percentage added to the team’s save percentage while a player is on ice. For example, if the team has an on-ice save percentage of 94% and a shooting percentage of 12% while Gabriel Bourque is on the ice, his PDO is thus 106. Historically, it’s been shown that for most players, there’s a regression toward a mean of 100 for this statistic. With that in mind, with a PDO of 106, Bourque should be considered “lucky” and his effectiveness most likely won’t sustain at that level. Sergei Kostitsyn may be the best recent example: for two seasons, he finished with a PDO of close to 105, mostly owing to his high shooting percentage. On the other side, if a player has a shooting percentage of 3% and an on-ice save percentage of 90 percent, his PDO is 93. History has shown that things will turn around for that player eventually, truly just a “down season.”  Looking back at some of those teams that could be labeled as “overachieving,”  you’ll see many high PDO numbers, indicating that “luck” was a factor.  Of course, nothing influences luck like good, hard work, but eventually, it can only carry you so far.

A lot of these statistics are reviled by some of the old-school types: Don Cherry, Phil Esposito, Harry Neale. To them, there’s no better means of measurement than the eyeball, and I don’t disagree. All of these metrics are intended to augment your analysis of the game on the ice, not replace it. While some of these “traditionalist” types will tell you that goals and assists still tell you more about a player than any Corsi rating ever could, there’s enough empirical evidence to suggest otherwise. At the very least, I would suggest combining these old standards with some of these new statistics I’ve discussed.


Extra Skater – Advanced Hockey Statistics

A relatively new site that I really like.  Allows you to take each individual game and review the advanced-stats results for each teamplayer.

Behind The Net

Gabe Desjardins (one of the godfathers of the advanced stats movement) maintains this site, long-considered the fancy-stats bible.


The Faker’s Guide to Advanced Statistics

Arctic Ice Hockey’s Guide to Advanced Statistics

NHL Advanced Stats Aren’t So Advanced, but they’re Better than Nothing – On the Forecheck

Photo Credit: Sarah Fuqua