-
-
by
Brent Hedtke
Some folks in the hockey community live and die by “fancy” stats and metrics, whereas others quite outwardly despise them. I tend to fall more in the former category but I certainly understand and empathize with the “watch the game” crowd. No one wants to feel like they’re doing a calculus exam while watching a game and advanced stats can certainly lead to the heavy eyelids that often precede deep REM sleep. Ultimately though, these stats are predictive and go a long way toward helping us determine what will happen as opposed to has.
NHL Advanced Stats
Corsi/Fenwick
Despite the goofy names, these are ostensibly the backbone of all advanced possession metrics. In essence, Corsi is just a simple plus/minus for shot attempts. That’s it. It measures the number of shots on goal, blocked shots and shots that miss the net. Over the course of a game, series or full season, a positive Corsi rating can be a great predictor of future success.
So let’s say over the first four games of a series a team has attempted 25 more shots than their opponent but has been outscored by five goals and is down three games to one. The positive Corsi rating is an indicator that they’re driving play as well as controlling possession, and likely just not getting the bounces. We can therefore predict that at their current rate of play, they are well positioned to turn things around in the remainder of the series, at least more so than a team with a negative Corsi rating would be.
Fenwick is almost identical to Corsi except that it doesn’t account for blocked shots. Most people prefer to use Corsi over Fenwick but there are many models that prefer the latter. Which leads us to the next, and arguably most instructive advanced stat of all.
Expected Goals
Not all shots are created equal. Corsi and Fenwick are great for assessing control of play but you’ve likely noticed that they don’t account for quality of shot. Let’s say Team A out-attempts their Team B by 25 shots to 10 over the course of a period. Attempting 25 out of the 35 total shots taken in a period will give you an excellent Corsi rating of almost 72%. But what if 20 of those 25 shots were from low-danger areas and all 10 of the opponents’ shots were from high-percentage scoring areas? In this scenario Corsi doesn’t do Team B justice. This is where Expected Goals come in.
Expected Goals For (xGF) and Expected Goals Against (xGA) are weighted metrics that consider shot type, location and circumstance to assess a value to how likely it is to result in a goal. Where it becomes useful in predicting future results is its ability to tell us what should be happening as opposed to what is. Let’s say over the course of a series Player A has an xGF of 3.8 goals but has only scored one. We can see that he is, for whatever reason, underperforming when it comes to finishing. This tells us that his raw goals total should ultimately catch up to his expected goals total.
PDO
At the end of the day, PDO ostensibly measures luck. The stat is compiled by adding up a team’s shooting and save percentages. All things being equal, this number should equal 100. Any variance above (lucky) or below (unlucky) 100 is an indicator that a team is getting the bounces or being punished by the hockey gods (unlucky).
Let’s say the league average save percentage is 91.2%. Simple math would then dictate that the league average shooting percentage would be 8.8%. Add them up and you get 100. If Team A is sitting on a league average save percentage of 91.2% but shooting at a 12.5% clip, this would result in an alarmingly high PDO of 103.7 and indicate that they’re due for a regression. That shooting percentage will likely come back down to earth and level out at league average. Conversely, if a talented team is sitting on a low range PDO of say 98.1, you could logically deduce that they’re simply not getting the bounces or suffering a bad stretch of goaltending.
Exceptionally talented teams can outperform their PDO and talent-deficient teams can underperform theirs but ultimately, everything regresses to the mean. PDO simply helps us tell which teams are most likely due for regression.
WOWY
With or Without You. It’s not just a U2 song. WOWY helps us measure which players on a line or tandem are actually driving play. If Player A has a Corsi rating of 55% when he’s away from Player B, but a 45% Corsi rating when together, we can deduce that Player B is “dragging” Player A down. Conversely, if Player B has a 45% Corsi rating when away from Player A, but a 50% Corsi rating when together, we can deduce that Player A is “carrying” Player B and more responsible for driving play.
It’s by no means a perfect stat, but where it can be instructive in a playoff setting is in the case of injuries or line adjustments. If we know that Defensemen A is being dragged down by his partner and a coach finally splits them up. We can predict a likely uptick in performance from Defenseman A. Same with injuries. The removal of a negative-impact player from the line-up can oftentimes be a net positive for the team as a whole.
Relative – This term measures a player’s on-ice performance relative to that of his teammates. Say that Player A has a Corsi rating of 51%. That’s good. But now say that the rest of his team is running at a 60% rate. All of sudden that 51% doesn’t look so impressive. A kissing cousin of WOWY, relative stats simply help us determine how a player is performing in relation to the rest of their team.
5v5
The truest value of any statistic is measured at even strength. 5v5 simply means five players and a goalie on both sides of the ice. Special teams can greatly alter a teams’ numbers and give a false impression of certain values. A team with elite shooters can juice their shooting percentages on the power play but that isn’t necessarily indicative of their overall success rate as the majority of the game is played at even strength.
5v5 measurements are most insightful when assessing goaltending performance. A goalie getting lit up on the penalty kill is more reflective of the performance of the team in front of them than their own, as goaltenders are at a massive disadvantage when killing penalties. The truest measurement of a goaltenders’ performance must be done at 5v5.
Per 60
Stats measured “per game” aren’t necessarily indicative of performance due to the simple fact that all players receive varying levels of ice time. To level these numbers out, we measure individual stats on a “Per 60 Minute” basis.
Say Player A scores a goal, two assists and records 12 shot attempts in 18:30 of ice time. Now say Player B scores a goal, an assist and records 8 shot attempts in 12:45 of ice time. Player A had better “per game” totals but when you account for ice time and extrapolate over a full 60 minutes of game time, Player A has significantly lower Per 60 totals, to the tune of 3.24 goals Per 60 to Player B’s 4.82, and so forth. So who is actually providing more value?
