There's been a lot of fall out since the Penguins lost game seven to the Rangers. Both Ray Shero and Dan Bylsma were fired, and the Penguins installed Jim Rutherford as the permanent-but-sort-of-temporary General Manager. Folks on twitter and in the media have also started crafting their narratives to explain what went wrong. A lot of people think that the Penguins, despite their strong possession numbers, didn't get quality chances and had to settle for relatively harmless shots. Examples are here, here, here and here.
My suspicion was that the Penguins got plenty of quality chances--roughly commensurate with their possession numbers--and that they lost primarily because of a low shooting percentage rather than poor strategy. To find out if this was the case, I went back and re-watched the entire NYR-PIT series and counted scoring chances.
I'm not doing this on a blank slate, though. The idea for this project came from Gabriel Desjardins more than four years ago, when he investigated similar claims that were made following the Penguins-Canadiens series in 2010. Gabe is the founder of Behind the Net, and reading his work the last five years has changed not only how I watch hockey, but also dramatically improved my understanding of the niceties and nuances of the game. His conclusion in that piece is always worth repeating:
Fundamentally, we have Jaroslav Halak playing better than his established talent level, Pittsburgh's shooters failing to make their shots at their career rates, and Montreal's shooters exceeded their own career numbers. Call it luck, call it personal performance highs - either way, it's not a persistent talent - it's something we all know might come crashing to the ground tomorrow. Of course, that means nothing is certain: nine times out of ten, Pittsburgh or Washington would be in the conference finals...This was time #10.
Before I present the data, a quick review of what I'm counting. I tracked all seven games in the Rangers-Penguins series for scoring chances, which are unblocked shots (shots on goal + missed shots) from within the "home plate" area of the ice depicted below (please click all images to enlarge them):
I didn't include blocked shots in scoring chances because blocked shots don't make it to the net to create scoring chances. In a sense, blocked shots tend to mean that the defense sort of did their job. But nevertheless, it doesn't matter if you track shots + missed shots or shots + missed shots + blocked shots as long as you're consistent with both teams. Here, we're only looking at shots on goal and missed shots.
I also only tracked even strength play (5v5 and 4v4) and excluded empty net data when one team had six skaters on the ice. I chose not to look at special teams because possession and scoring chances aren't as illuminating in those circumstances. No one's impressed you out-possessed or out-chanced the penalty kill unit because you had an extra skater on the ice. Scoring chances are most useful when measuring even strength play, as they indicate who carried play when the playing field was level. Trying to figure out what went wrong with Pittsburgh's PP in this series will require different metrics, and I hope to take a look at that in a later piece.
Scoring Chances by Individual Player
Below is a chart with the scoring chances for each player. A quick note on abbreviations. SC = scoring chances, or the number of scoring chances that individual player had; SCF = scoring chances for, or the number of scoring chances for the Penguins while that player was on the ice; SCA = scoring chances against, or the number of scoring chances against the Penguins while that player was on the ice; SC +/- = scoring chance plus-minus, which is calculated by subtracting scoring chances against from scoring chances for; and SCF% = scoring chance for %, or the percentage of scoring chances the Penguins had while that player was on the ice.
Overall, this is dominant stuff. The green boxes are (obviously) the guys with a positive scoring chance plus-minus and a SCF% above 50%. Red, not so much. I'll bypass Orpik and Glass in discussing this because neither played that much, but it's a bit comical (to me at least) how bad Tanner Glass is.
As for the defensemen, three things in particular stood out to me. First, it's impossible to miss just how insane the Martin-Letang pairing was. They skated about 27 minutes a night against the best players on the Rangers and came out way ahead in terms of scoring chances. The craziest thing is that Martin was on the ice for 50 Penguins scoring chances even though the Pens only had 82 as a team. That means Martin was on the ice for 61 percent of the scoring chances in Pittsburgh's favor despite only playing about 45% of each game.
The second thing that struck me was how disappointed some team is going to be in December when they realize Niskanen isn't worth the $6 or 7 million dollars per year that they're paying him. I thought his performance tailed off significantly as the season wore on, and while the playoffs are a small sample, it's alarming that he couldn't be better than Scuderi or Bortuzzo despite his cushy minutes.
Which brings me to the third point. Though Scuderi is awful and Bortuzzo isn't much better, they were both positive scoring chance players. It seems unbelievable at first blush, but a big reason for this is because Bylsma did an excellent job of sheltering Scuderi in the Rangers series. He saw easy competition and zone starts typically above 55%. The upshot is that Scuderi was (surprisingly) not a major reason the Penguins lost to the Rangers.
As for the forwards, it's more of the same. Malkin was a beast, which underscores the utter stupidity of the trade Malkin propositions. And that Sutter-Neal-Jokinen line was quietly very effective. I think the numbers were already there for Jussi during the regular season, but these only reinforce the message: sign this man to a reasonable deal. He's a very good hockey player who generally makes everyone better.
Even the bottom six plugs who take a constant beating made sure that the majority of scoring chances were directed at Lundqvist. In particular, I think that Stempniak flies under the radar a bit but his stats are quite good. Despite generally getting bottom six minutes, he had more scoring chances than Jokinen and nearly as many as Crosby. The Penguins need to ink that man to a reasonable deal before free agency starts.
In darker news, Beau Bennett. Woof. I thought this kid would break through this year, but he was so disappointing in the playoffs. He wasn't terrible, but he was the only forward in red on both counts, despite playing on a bottom six unit that's supposedly thin. What hurts even more is that he's supposed to be our future of young, skilled, fast hockey players. I won't belabor the point because his injury probably held him back, but he stunk it up this year in the month of May.
Explaining Crosby's "Struggles"
Among the 243 skaters who played at least six playoff games this year, Crosby was ranked second in CF% at almost 62%. Given how important possession is to winning in the NHL, it's reasonable to look at that and think that Crosby was a lethal hockey machine during these playoffs. And in a sense, you'd be right. Crosby kept the puck in the offensive zone more than almost anyone else, despite coaches specifically game-planning to put their best players on him in order to shut him down. The fact that he still had the puck so much is, in a sense, breathtaking.
But he only scored one goal in 13 games. And my memory of the Penguins games from this year's playoffs led me to think that Crosby wasn't as dominant as those numbers made him out to be. Shooting percentage is so variable though; it's not unheard of for goal scorers to go very cold over 13-game sets. Alex Ovechkin led the league in goals despite not scoring an even strength goal for more than an entire month.
I think the scoring chance data presented above provides the bridge between these competing storylines that helps us make sense of Crosby's postseason. While Crosby had 62% of the Corsi events on the ice, he had far less control over the scoring chances (only about 56%). And that difference of six percentage points is huge given that most players are clustered between 40% and 60%. What this illustrates is that while Crosby was exceptional at maintaining puck possession, he was less dynamic when it came to out-playing the other team for quality scoring chances. Indeed, Crosby was on the ice for the second-most chances against of anyone on the team. Paul Martin saw less total scoring chances against despite playing nearly six more minutes a night than Crosby.
The benefit of this, in my opinion, is that we can make this claim about Crosby's play without relying on an "eye test" that frequently leads us astray or our memories which can't store all of the bits of information produced by NHL games. The scoring chance data is an objective, measured look at whether Crosby was as elite as his possession numbers suggested. He was not. Why that is so is pretty difficult to answer without inside information. But the most salient point is that we don't need to latch onto fuzzy recollections or random fluctuations in shooting percentage to analyze this question. Watching and tracking the games will do just fine.
Scoring Chances at the Team Level
I also tracked scoring chances on the team level, and divided them up game-by-game. The chart below contains the relevant information.
|Game 1||Game 2||Game 3||Game 4||Game 5||Game 6||Game 7||Total||SC +/-||SCF%|
This is (of course) in line with the data for individual players. The Rangers out-chanced the Penguins in only two games; one was by the thinnest of margins, and the other (game 3) was when the Penguins had a 2-0 lead in the third period and succumbed to score effects. And this is what we'd expect from looking at the possession numbers. The Penguins dominated the Rangers with a 56% FF in score-close situations during the series. As Eric Tulsky concluded in an article on shot quality:
Whatever tendency certain players might have for driving their team to get more scoring chances than a simple shot differential predicts is small and swamped by random noise. This suggests tracking scoring chances isn't adding much information to the readily available shot differential numbers.
Eric has noted elsewhere that Corsi and scoring chances converge in the long-run, so they may not be as tightly linked in a small sample like this seven-game series. But the fact that the Penguins' SCF% was only a few percentage points higher than their fenwick close is reassuring. I'd be skeptical if someone tracked these game and got a scoring chance count that was significantly different than the possession numbers.
For more visual folks, here's a graphical representation of the scoring chances for each game.
It's remarkable how severely the Penguins brutalized the Rangers in games 2 and 4. And in case anyone thinks the data from those games is skewing the overall results, the Penguins were still pretty darn good in the games they lost.
You can see that the Penguins out-chanced the Rangers in three of the four losses, and they especially pulled away in Game 7. The fact that the Penguins had five more scoring chances than the Rangers in that game--and lost--is pretty startling. Remember: the Rangers only had 4 even strength scoring chances in all of game 2. Said another way, the Penguins had 61% of all scoring chances in game 7. If they put those numbers up routinely, they'd be elite.
But the Penguins still lost. If you go to extra skater, you'll see that at even strength, the Rangers and the Penguins each scored 11 non-empty net goals during this series. But with the disparity in scoring chances, that means Lundqvist and Fleury were far apart in save percentage on scoring chances. Not every goal the Penguins scored at even strength was a scoring chance, but let's assume they were (which only makes Henrik look worse than he really was). His save percentage on scoring chances would be 86.6%. For Fleury, one of the goals scored by New York was not a scoring chance, which leaves ten even strength goals. That means Fleury had a scoring chance save percentage of 83%, worse than Henrik's numbers even after we've made Henrik look bad by assuming all his goals were scoring chances and removing one from Fleury's data set.
That difference of 3.6% might not seem like a lot, but 3.6% of the 82 scoring chances Pittsburgh had equates to 3 goals. That's huge, especially considering we only needed one goal to win game 1 and two goals to win game 7.
Tracking these games and writing this post was cathartic, refreshing, and depressing all at the same time. I wanted to track scoring chances to see if the Penguins were part of an incredibly rare script that saw an NHL team in today's league dominate possession but fall victim to a neutral zone trap of sorts that led to an overwhelming majority of harmless shots. That wasn't the case. The Penguins had significantly more high-quality chances than the Rangers, and it isn't skewed by the data in the games Pittsburgh won.
Which is why this exercise was depressing. I'd be inclined to agree that something was fundamentally wrong with the system or the gameplan if the Penguins couldn't generate possession, or if their scoring chance numbers were far, far lower than their Corsi or Fenwick numbers. But since that wasn't true, I'm not sure what could have been done differently. Of course, they could have hypothetically played better, but they already put very good numbers up against a very good Rangers team.
Others are likely tired of hearing the "hot goalie" excuse, but a sober review of the Rangers series, in my opinion, points in only that direction. In a game driven by probabilities, a seemingly inexplicable sequence of events can happen. The key is to understand what you can and can't control, and to react accordingly. Overreactions are never good, and I hope this isn't lost on the men in our front office.