Nashville Predators: A Crash Course in Modern Hockey Analytics

NHL teams and media are transitioning to a modern understanding of the game. Using the Nashville Predators as an example, let’s break it all down.

To slightly oversimplify the debate, there are essentially two schools of thought when analyzing hockey. The first involves the so-called “eye test.” Basically, this perspective values points on the board and visually apparent contributions of a player. The alternative to this involves a deeper usage of analytics. With some examples from the Nashville Predators, here’s a breakdown of the most commonly used “advanced analytics.”

Often, only statistics at five-on-five are included, omitting special teams situations. Special teams inflate certain metrics, and can paint an inaccurate picture of a team or player’s performance.

The goal of these statistics is quite simple. Teams want to dress their best lineup for every game. To do so requires a great deal of understanding as to the specific contributions of each player. Hockey is a fast-paced, nuanced game. The more comprehensive a metric, the more helpful it will be in making those tough decisions.

Corsi & Fenwick

Also referred to as “shot attempts (SAT),” Corsi is one of the more basic analytical tools in hockey. Basically, it is an extension of the common “shots on goal” metric. Corsi was originally created by Tim Barnes, a financial analyst from Chicago. Since its creation in 2007, it has become easily the most ubiquitous “fancy” stat in hockey.

Corsi’s strength is in its simplicity. It does not make any assumptions or require a high degree of calculation. A team’s “Corsi For” is the number of shot attempts they create, while “Corsi Against” refers to those attempts created by the opposing team.

At its root (core, if you will), Corsi is a proxy for a team’s possession. The higher the proportion of shot attempts created vs those allowed (Corsi For %), the more time a team spent in the offensive zone.

Fenwick is nearly identical, but refers only to unblocked shot attempts (USAT). The argument here is that a team should not get offensive credit for selecting a closed shooting lane. I prefer to use Fenwick in my analysis because, in my opinion, it represents the game flow more accurately.

Kyle Turris is a perfect example of why this is a valuable statistic. Turris definitely puts himself on score sheets with goals and assists, but his impact in a game is so much more. Corsi/Fenwick allows you to analyze the Predators performance as a whole with players like Turris on the ice. In doing so, you can really tell how his presence affects the game overall, even without a goal being scored at either end.

Scoring Chances

Shot attempt metrics are great at helping you understand the general flow of a game, but aren’t able to really describe its specific nature. Scoring chances help us understand a team’s performance based on where each shot attempt occurred. If you’ve read any of my articles, chances are you’ve seen the following graphic from Andrew Berkshire of Sportsnet:

While Corsi counts all shot attempts equally, scoring chances allow some nuance. As Berkshire points out, a “scoring chance” must come from the “home plate” area, or dark gray and red portions, of the ice. This is entirely based on expected save percentage, on average. Taking a shot from the blue line is effective about 2% of the time, while a shot from the faceoff circle can jump to about 13%.

Scoring chances for (SCF) and against (SCA) provide the scoring chance for percentage (SCF%). This is simply the proportion of scoring chances the selected team generates throughout a game.

High-danger scoring chances are a drum that I beat quite often. These chances come from the red area of the diagram, where a goaltender makes the save roughly 77% of the time. If your team can effectively pressure the low slot, while defending its own, the result will likely be in your favor.

Roman Josi is a great example for high-danger chances. Throughout this season, he has helped generate 66 high-danger chances at five-on-five, while allowing 52. His speed, puck movement, and skating ability combine to make him one of the league’s top attacking defensemen. He has a great understanding of the game, and will always try to get the puck to a high-danger area. While paired with Mattias Ekholm, a much more conservative defenseman, Josi can be a truly dominant player.

Zone Starts

It’s not always fair to compare players directly, while assuming their usage is identical. Not all skaters are created, or used, equally. Tracking zone starts allows you to understand the specific role of a player throughout a game, season, or even career. This metric is based on faceoff location, either in the defensive, neutral, or offensive zones.

Certain lines tend to be “sheltered,” or allowed to be more offensive. For the Nashville Predators, this is true of the JoFA line. Their offensive prowess is maximized by limiting the amount of defending required.

Calle Jarnkrok provides a perfect example. If Jankrok’s Corsi numbers are compared to, say, Kevin Fiala‘s, you get significantly different values. This season, at five-on-five, Jarnkrok’s CF% is 43.54%, meaning that opposing teams generate more shot attempts than the Predators while he is on the ice. Fiala’s, on the other hand, is 51.69%. So, Kevin Fiala is the better forward, right?

Not so fast! Calle Jarnkrok is a specific type of player: a defensive forward. His line’s role is to suppress an opponent’s top players, rather than score. Take a look at their numbers, including games played (GP), Corsi-for percentage (CF%), and offensive zone start percentage (OZS%).

It’s obvious that Fiala generates proportionally more offense than Jarnkrok at five-on-five. However, Jarnkrok begins most of his shifts with a defensive-zone faceoff, which explains a major element in the discrepancy.

Score-and-Venue Adjustment

If the goal of advanced analytics is to better describe performances, it’s vital to inject some context into each metric. Score-and-venue adjustment (SVA) provides just that. When paired with another metric, say Corsi, it allows you to analyze a player or team during specific game situations.

Calculating score adjustments is based on the assumption that a losing team will play more offensively. This can skew shot production both ways, since their defensive coverage will likely suffer as well. By quantifying that trend, analysts have created different systems for bringing everything back to average.

Venue adjustments rely on similar assumptions. On average, home teams have generated more shot attempts than away teams over the course of several years. This average difference allows for an adjustment to be made. Instead of pretending that the venue has no affect on shot attempts, a venue adjustment can be made to mitigate the effects of that inherent disparity.

Take a look at the game flow from Monday’s game against Boston:

Switch between “Corsi 5v5” and “Corsi 5v5 SVA” and you’ll see what I’m talking about. The unadjusted 5v5 Corsi is a bad look for the Nashville Predators; the shot production moved more and more in Boston’s favor as the game progressed. However, by omitting the inherent difference in a team’s strategy when tied and when losing, the trend is much more neutral.

PDO

You might recognize PDO from Dimitri Filipovic’s excellent Hockey PDOcast. It’s the first metric I’ve brought up that actually combines two different types of statistics. Officially, the NHL refers to PDO as shooting percentage + save percentage, or “SPSV%.”

As the official name implies, this metric first employs a player’s shooting percentage, or the number of goals scored divided by shots on goal. That percentage is then aggregated with a team’s save percentage while the selected player is on the ice. It seems complicated, but the goal is rather simple: to accurately describe a player’s overall influence on his/her team’s performance.

I’ll be honest, I don’t find too much value in PDO. The average score is 100, and the rule is that everyone always regresses to the mean. The best example is combining a shooting percentage of around 9%, and a save percentage of around 91%. The result is 100, which is the mean.

A good example for the Nashville Predators is Cody McLeod. Currently, his five-on-five PDO is 101.5, good for sixth-best on the Nashville Predators’ roster. Is Cody McLeod the sixth-best player in Nashville? We would be in bad shape if that were the case.

The Predators’ shooting percentage with McLeod on the ice is 10.26%, slightly above league average. The team’s save percentage during his shifts is 92.3%, also above average. Everyone in the league understands the true role of Cody McLeod. He attempts to protect his teammates by use of force, not put them ahead on the scoreboard.

PDO is perfectly acceptable as a descriptor, but requires a bit more context than other metrics. For that reason, I tend to leave it out of my analyses.