With the series between the Penguins and Islanders wrapped up, both the mainstream and the Internet media begin the annual process of constructing narratives to describe what happened. There are hundreds of opinions out there explaining why the Penguins won, so it's risky to try and highlight what the dominant or even most oft-repeated themes are. But for this series, I think it's safe to say that the major theme is that the Penguins got severely outplayed and are lucky to be advancing. A caption for a post over at Faceoff Factor sums up that sentiment:
I think there are two problems with this narrative. First, it exaggerates what happened. Second, it ignored the critical influence of score effects.
I'll begin with my first complaint. The Islanders did outplay the Penguins for periods of time this series; Game 6 is the most recent and best example. But the Islanders possession advantage at even strength wasn't that big. If you look at the Time on Ice reports for Game 1, 2, 3, 4, 5, and 6, you can see that the Penguins had 196 total shots on goal and missed shots this series, while the Islanders had 232 shots on goal and missed shots over these six games. These numbers are each team's Fenwick score for the series (Fenwick is missed shots + shots on goal). To determine the percent of possession for each team, simply divide that team's score by the total Fenwick events for both teams. So for the Pens, they had a Fenwick of 196/(196+232) = 45.8%. By definition, the Islanders thus had a Fenwick of 54.2%.
This difference between the two teams was not large enough, in my eyes, to suggest that one team was dominated. Of course, we get into semantics by trying to figure out what exactly it means for one team to dominate another, but I think in a scenario where one team really controlled the play for most of the time, the dominating team's Fenwick should be at least north of 60%. And of course, it's important to note that we're discussing this right now without the critical context of score effects (something that shouldn't be done and that I'll discuss below). But the point is that even stripped of its context, the Islanders didn't tilt possession so much in their favor over this series to actually indicate they severely outplayed the Penguins.
Scoring chances tell a similar story. Here are the scoring chances for games 1, 2, 3, 4, 5, and 6. For those who are unaware, a scoring chance is generally defined as a shot directed toward the net from this area. At the end of the series, the Islanders led the Penguins in total scoring chances 105-96, and in even strength scoring chances 83-73 (remember, we're still ignoring score effects). This is an edge for the Isles, but it's not an extreme one, and certainly not one that suggests the Isles dominated/severely-outplayed the Penguins. An example of domination would be when the Penguins played the Canadiens in the playoffs in 2010 and led in scoring chances 159-89.
So all in all, there really isn't any evidence that the Penguins should consider themselves "lucky" to be moving on. Even without accounting for score effects, the Islanders were just slightly better than the Penguins at possession and scoring chances. Perhaps some might be upset the Penguins couldn't outright dominate the Islanders because they were supposedly a very bad team. But that was a fantasy created by the media that had no basis in reality. The Isles finished with 55 points in the standings; the third place Capitals had 57. No one would be complaining if they realized the Isles had as much talent as the third place team in the conference. What's more, the Isles played the penguins very tough the whole year, and they were also a better possession team than the Penguins by the end of the season. People should realize that the Pens beat a very good team in six games.
But the claim that the Isles were slightly better hinges on looking at Fenwick and scoring chances without the aid of score effects. This is the wrong thing to do, and I'm surprised that no one has mentioned it yet.
Score effects is a concept that describes how teams play to the score. Teams that are trailing out-possess their opponent, and consequently, teams that are leading tend to sit back and protect the lead as the game wears on. Score effects have been talked about for some time, and it's been verified across a number of seasons in the NHL. One example will show why using score effects is critical. Imagine, in a game, that team A dominates team B in possession and jumps out to a 3-0 lead at the end of the second period. Score effects would tell us that team B is going to come out and play some fast, desperate hockey in the third. They'll put a lot of rubber on net; their d-men will pinch in the offensive zone; and they'll probably draw some penalties. They lose 3-1 but because of their frantic play in the third, shots, scoring chances and Fenwick were even. Would we say team B was equally good but simply got unlucky? Of course not. Every team losing by three goals will dominate in the third, but that doesn't make up for the two bad periods of hockey in the first and the second. Without reading this game in the context of score effects, we'd erroneously conclude this was an even game, when nothing could be further from the truth.
As the narratives about the Pens-Isles series made their way onto the Internets Saturday night, I thought to myself: didn't the Isles spend a lot of time trailing in this series? To find out, I added up the totals for the different game states in this series. There were eight game states in all: score tied, Pens up by 1, Pens up by 2, Pens up by 3, Pens up by 4, Pens up by 5, Isles up by 1, and Isles up by 2. The chart below shows the amount of time in each game state:
So it looks like I wasn't crazy: the Isles did spend a ton of time behind on the scoreboard. Out of the roughly 376 minutes played in this series, 43% of those minutes were played while the Islanders were trailing. In contrast, the Penguins only trailed about 15% of the time. Said one way, the Isles were trailing in this series nearly three times as often as the Penguins. Said another way, the Isles were trailing so often that it was as if they had played eight consecutive periods of hockey while down by at least one goal. The bottom line is that with all this time trailing, the Islanders were definitely going to out possess the Penguins. So what happened in this series is completely expected. In fact, I think that given all the time they spent trailing, the fact that they were only able to rack up a small advantage indicates the Penguins actually were the slightly better team this series.
I know, your head just exploded. Let me explain.
Aside from simply saying "the Isles trailed for much of the series, so thats a big strike against them," I wanted to look at each team's Fenwick score for the game states where the Isles trailed. So I went into the play-by-play reports for games 1, 2, 3, 4, 5, and 6 and added up the Fenwick for each team at those game states. With those numbers, I could figure out what percentage of possession the Pens controlled at each game state, and then compare that to the league average.
But before I get to the chart, here's a couple of caveats:
- I only focus on Fenwick because there is no reliable score effects data for Corsi (or at least none that I know of).
- What do I mean by league average? I mean, for example, what the average team's Fenwick is when they're trailing by one goal, two goals, etc. Eric T. wrote a great article which highlighted that the average Fenwick for a team that is trailing by two goals is 56% (and thus, the average Fenwick for a team leading by two goals is 44%). For teams trailing by one goal, the average Fenwick is 54%. Behind the Net, however, doesn't collect reliable data for teams that are leading/trailing by 3 goals or more. This article did collect those numbers, however, for the 16 playoff teams in 2009-10. This is hardly a representative sample, but it shows that the average team here, when trailing by 3 goals, got 57% of the possession. I'm going to use that as an average in my sample.
- There really is no data at all for teams that are trailing/leading by four goals or more, probably because these game states are fairly uncommon. In addition, while the Isles did spend a lot of time trailing by four goals or more, there were very few Fenwick events during this time, so it's unlikely these numbers would represent a signal rather than random noise. Therefore, I disregarded them.
- The tough thing about score effects (especially in the third) is that they're dynamic rather than constant. What I mean by that is that it's not true that a team trailing by 2 goals in the third gets 56% of the possession the whole 20 minutes. Instead, score effects get more pronounced as the game wears on, so much that during the last three minutes or so, the trailing team will likely have total possession. The graph in this post puts this in pictorial form. However, to make matters easy, I'll assume score effects are constant. And to be clear, this helps the Islanders. They were trailing a lot more in the third than the Penguins; 54:37 min (Isles) compared to 29:39 min (Pens) to be exact. So they'd be expected to make an even more desperate push that won't be captured in my analysis.
- Finally, this is only focusing on even strength events. But I think it's important to remember that the Penguins clearly won the special teams battle in this series. Whether pundits or cynical Penguins fans like to admit it, special teams are as much a part of the six game series as even strength play, and the Penguins (and coaches) deserve plenty of praise for shutting down a dangerous Isles PP and maintaining a lethal PP of their own.
As I referenced above, the Isles spent much more time trailing, so I focused on those game states specifically because we want to give the most attention to (or weight the most heavily) the game states which occurred the most.
Three things stick out to me from this table. First, the two teams were very close when the score was tied. Though the Isles had a 51.8% advantage in Fenwick with the score tied, the situation could easily have been reversed. Why? I had 199 total Fenwick events when the score was tied: 96 for the Pens and 103 for the Isles. Imagine if just one shift was different; not one game or one period, or even one half of a period. Just one shift. That is, instead of the Isles directing shots at our net during a 40 second span, imagine the Penguins dominating zone time and throwing two shots on goal and two more missed shots. This would have been a great shift, but still only one shift out of hundreds. But that's a swing of four points, and would give the Pens a 50.3% Fenwick edge (100/(100 + 99)). This hypothetical is only meant to illustrate how tight the play was at even strength. Indeed, the Pens and Isles played literally even hockey over the course of six games. Yes it ebbed and flowed in either team's favor, but I think it's a mistake to say one team was outplayed.
Second, the Pens were average when it came to playing with a two and three goal lead. This again reinforces the notion that the Penguins were not outplayed. Finally, and most importantly, the Penguins were well above average when it came to playing with a one goal lead.
In the end, I think the Penguins slightly outplayed the Isles given their performance while the Isles were trailing. They were average when up by 2 and 3, they were basically even with the score tied (an average performance), and they definitely outperformed when up by one goal. I do want to caution that I think any edge for the Pens isn't that big; their performance with the score tied is the best indication this was a pretty even series, and the ultimate Isles edge in scoring chances and possession at even strength was entirely driven by score effects.
Hopefully the Penguins play better against Ottawa, but they definitely deserved to be moving on.