Hockey Analytics: dissecting the 2011-2012 season, piece by piece


Hi, I’m Mathieu Roy, I’m a Habs fan, and I’m into hockey analytics. (Hi Mat!) I hope to bring stories with an analytic bent to A Winning Habit; and since it is the off-season, I thought I’d start with a season review.

It’s customary to think of a hockey season in terms of segments. Typically, 10-game segments are used, but as I was looking over the data for the event-filled 2011-2012 season, I wondered if some of those events wouldn’t make for better cut-off points. So I split the season off in different sections, often but not always bordering them with some of the more significant happenings during this season, resulting in nine segments ranging from seven to twelve games. Because of the somewhat arbitrary lengths of the segments, this chart is going to be more illustrative than scientific, but it will nonetheless give a good idea of how the fortunes of the team changed over time. I also hope to get around to doing this exercise for specific players, comparing them to the overall team (keeping in mind the different roles of the players).

For our readers who aren’t familiar with hockey statistical analysis, this piece mainly uses three metrics, all of which are closely related together.

The blue line on the chart represents the ratio of scoring chances as counted by Olivier. This is defined as a hard play on net from the “home plate area”, covering the area from the goalposts to the faceoff dots, from there to the top of the faceoff circles; blocked shots are never counted. More intuitively, they represent shot attempts from high-percentage, non-perimeter areas.

The yellow line represents the team’s Fenwick metric over the time period. This is a sum of all goals, shots and missed shots relative to the opposition’s sum of the same. Again, blocked shots don’t factor in. Fenwick is a proxy for puck possession, helping measure which team has the puck more during a game. It also tends to track closely with scoring chances, which we certainly see in this chart.

The green line is Fenwick with the score “close”, either tied, or within a goal in the first two periods. This is important because teams change their strategies when they play to the score. Leading teams tend to get outshot by trailing teams. Close-score situations help eliminate these score effects from the metrics, and also give us an idea of how the team performs in these high-leverage situations.

All of them are expressed as a percentage, which represents the Habs’ share of the total; so if they outchanced the other guys 12-8, they’d have 12 out of 20 scoring chances, or 60%. I drew a flat red line for the 50% bar.

Also, it is important to note that *all these stats cover 5-on-5 exclusively*. Special teams are a different can of worms, one that we might get into in the future.

For this piece, I’m indebted to Olivier Bouchard, who assembled the data I’m massaging here, notably by painstakingly counting scoring chances for all Habs games. Olivier was profiled on A Winning Habit recently and as a Francophone, I cannot recommend his blog enough; this is really his data, I’m just wrangling it.

With that acknowledgement, let’s have the chart, followed by a quick segment-by-segment analysis of the season. The X-axis is properly scaled on a per-game basis, and the final value for each metric is that of the entire segment in question.

 

 

SEGMENT 1: October 6 to October 24: 8 games, 1-5-2

Chances: 53.65%, Fenwick: 52.74% Fenwick Close: 52.96%

This segment goes from opening night to just before Perry Pearns is fired, ostensibly because of the floundering power play. From the chart we can see that the Habs actually perform mostly excellent at even-strength, a trend that is reflected by their goals. That’s despite a fairly significant injury situation: Campoli is infamously injured on opening night, Cammalleri and Spacek are both lost on October 9, and when Cammy returns on October 20th, Gomez gets injured. At this point, analytics buffs are actually seeing good things in the Habs’ future: the 5-on-5 is solid, and while this hasn’t translated into wins, it seems to be mostly a matter of time.

 

SEGMENT 2: Octboer 26 to November 10: 7 games, 5-2-0

Chances: 49.70%, Fenwick: 52.73%, Fenwick Close: 54.72%

This is an excellent segment for the Habs; they finally start getting rewarded for consistently excellent play. It’s unusual that the scoring chance metric is below 50%, so far away from the Fenwick curves; however, this quirk does not impact their record.

This segment goes from the Pearns firing to just before November 12; this happens to be halfway to the end of the next segment, but it also coincidentally falls just before the game where Gomez would return to the ice and Cammalleri would get injured. Andrei Kostitsyn gets injured on November 8, too, continuing the Habs

 

SEGMENT 3: November 12 to November 23: 7 games, 4-2-1

Chances: 54.27%, Fenwick: 54.15%, Fenwick Close: 54.24%

This was possibly the best segment of the year for the Habs. They are dominating at even-strength, they’re winning games and looking for all the world like the 100+ point team they were supposed to be, and look poised to overcome their weak start. Keep this up for the rest of the season, the playoffs are in the bag and the Habs look at least like a dark-horse Cup contender, especially if Markov comes back in anything resembling his usual form. This isn’t hyperbole; at the quarter-pole of the season, the Habs are looking like the excellent even-strength club they haven’t been in at least a decade. And they’re doing this minus Markov and Kostitsyn, and with Hal Gill missing most of this segment due to a virus.

On November 23rd, against Carolina, Gomez was injured once again. Coincidentally or not, that was when the bottom completely fell off on the Habs’ season.

 

SEGMENT 4: November 25 to December 15: 10 games, 3-3-4

Chances: 47.11%, Fenwick: 45.54%, Fenwick Close: 42.90%

This was the segment that would cost Jacques Martin his job.

It’s become fashionable to pin all the woes of the world on Gauthier and, certainly, his in-season personnel work during the season was execrable. His firing of Martin on the basis of a ten-game span was certainly too hasty. But when he said, after firing Martin, that the team hadn’t been playing well, he certainly wasn’t wrong. Whether the coach was at fault was another matter entirely. The defining

First was Gomez. For all his faults, real or imagined, he had heretofore been very strong at driving the play towards the opposing net. But he certainly cannot, by himself, explain why a team that owed the puck 54% of the time suddenly fell to 45%. Combined with Markov still being out, and losing Gionta on December 8th, then Moen on the 13th… Gomez’s second injury may simply have been the tipping point, the time where they simply ran out of people to use to patch over the holes that the mounting injury situation kept punching into their boat.

Trading Spacek for Kaberle was no help at all. Kaberle was a power play specialist for a team who was starting to flounder at even-strength and, really, was having no trouble generating chances on the power play. Of the two, Spacek was more likely to be able to help.

On December 17, in a futile attempt to save a season gone wrong Jacques Martin was fired and replaced by Randy Cunneyworth. This was a mistake, because Martin didn’t deserve to be fired because of a 10-game span in an otherwise excellent but injury-plagued season, and because Cunneyworth would prove not to be up to the task of coaching a NHL club.

The bottom had fallen before Martin was fired. But this removed any hope the Habs had of recovering.

 

SEGMENT 5: December 17 to January 14: 12 games, 3-8-1

Chances: 50.18%, Fenwick: 47.44%, Fenwick Close: 45.54%

One may be tempted, looking at the graph, to conclude that the Habs derived immediate dividends from the arrival of Cunneyworth. A quick glance at the record over this segment should disabuse the onlooker of this notion, and the still-abysmal Fenwick Close score should clinch it.

The Habs improved because, frankly, they had nowhere to go but up. 42.90% close Fenwick was unconscionably low, and it was basically impossible for a team to maintain this level of sucktitude for long. Sheer regression to the mean would’ve carried the Habs up, and at 45.54% with the score close, the Habs continued to be utterly terrible. Score effects contributed to make the rest of their stats look better; they ended up trailing a lot, and spent a lot of time facing teams that sat on leads.

This segment contained a number of interesting events as Cunneyworth felt his oats. He struggled with figuring out who was his best tough-minutes center (before finally, and unsurprisingly, settling on Plekanec) and figuring out whether Eller was a first-, second-, third-, fourth- line center. Or whether he belonged in the lineup at all; Eller was inexplicably scratched on January 4th. More inexplicably, Cunneyworth also scratched his hands-down best defenseman in Subban (if not best skater outright) in that same game, at a time where the Habs had Gionta, Gomez and Markov already on IR.

Before this stretch, the playoffs were still a credible objective; they’d managed to tread water in the standings during segment 4, and still had a realistic shot at making the playoffs. This was the stretch that put the Habs definitely out of playoff contention.

Ironically, this stretch featured only one major injury: Gionta returned briefly on the 7th, only to leave the game on the 10th, this time for good. Olivier likened this event to a man hitting the ground after falling 50 stories, telling himself “so far, all is well” all the way down until the inevitable splat.

This was when the Habs went splat. To emphasize that the lights had gone out, Gauthier traded Cammalleri away, in the middle of a game, on January 12th. In return, the Habs obtained Rene Bourque, whose suspension would only expire during the next segment, and who would prove to be a far inferior player.

 

SEGMENT 6: January 15th to February 5th: 9 games, 4-4-1

Chances: 52.02%, Fenwick: 49.01%, Fenwick Close: 46.82%

This segment marks the arrival of Rene Bourque and the second return of Scott Gomez. An interesting segment, because the chances don’t follow puck possession at all. The Habs parlayed that into a less disastrous record than the previous sequence, though the playoffs had long become a pipe dream. The Habs continue to lag far behind their opponents on puck possession; despite this, worse was yet to come.

 

SEGMENT 7: February 7th to February 26th: 10 games, 4-5-1

Chances: 45.85%, Fenwick: 44.28%, Fenwick Close: 42.16%

Statistically, the 10-game stretch leading to the trade deadline was, hands-down, the worst the Habs had all year, managing to outdo even the 10-game segment preceding Jacques Martin’s firing.

The demolition of the team was concluded during this stretch as management, faced with the certainty of a lost year, traded away expiring contracts at the deadline. Gomez, who may have still suffering from his injury, was moved to the fourth line. The alienation of Kostitsyn continued, culminating to his trade to Nashville at the deadline, following Gill by a few days.

 

SEGMENT 8: February 28th to March 16th: 9 games, 4-3-2

Chances: 47.01%, Fenwick: 43.01%, Fenwick Close: 42.57%

Another stretch as bad as the infamous Segment 4. But at this point, between the firesale and the injuries, the Habs simply did not have the horses to be competitive. They were simply playing out the string to see where the lottery chips would fall.

One bit of good news: Markov did finally return to play on March 10th, and even though he looked like a guy who hadn’t played in more than a year, there nonetheless was much rejoicing.

 

SEGMENT 9: March 17th to April 7th: 10 games, 3-3-4

Chances: 48.83%, Fenwick: 44.61%, Fenwick Close: 45.63%

The final stretch of the season is barely any better than the previous one, but during that time the Habs manage to lose Carey Price, Travis Moen, Mathieu Darche, Ryan White, and Scott Gomez, all to concussions, putting the exclamation point on a lost season, and landing them dead last in the Eastern Conference (despite a differential of merely -2 in non-empty net goals).

 

CONCLUSION: The 2011-2012 Montreal Canadiens were a case study in how injuries, transient problems such as power play finish and goal distribution, and panicky management can derail a team on a promising run. Up until the quarter pole, the Habs were one of the best teams in the league at even-strength, something which they hadn’t been in a decade. Unfortunately, bad shooting luck on the power play and injuries led to a record that was not representative of the team’s level of play, leading to the front office making hasty, destructive personnel decision both behind and on the bench. The team that finished the season on April 7th wasn’t even a pale imitation of that early-season club; it was little more than a shipwreck.

The sudden drop between those 21 early games and the other three quarters can hardly be overstated. In that first quarter-season, the Habs got 52.63% of scoring chances, 52.90% of Fenwick events, and a whopping 54.5% of Fenwick events with the score close. The rest of the way, they were 49.43% on scoring chance, 47.73% on Fenwick, and 43.94% on score-close Fenwick. To give a sense of scale, over the course of the season a score-close Fenwick of 54.5% would have been third-best in the league; 43.94% would have been dead last, and the Habs were consistently that bad through the last three quarters of the season.

Can 2012-2013 recapture the strength of that first quarter? This seems doubtful, due to one problem, that of personnel, and one question mark, that of coaching.

First, the Habs lost Cammalleri and Kostitsyn to trade, and neither of them were really replaced (in a perfect world, Bourque might replace Kostitsyn, but he appears to be a worse player). The Habs have added to the bottom of their lineup, but that only ever became a problem because the top got decimated and the bottom-sixers ended up having to go up the depth chart, forcing the team to use AHL journeymen to replace them. While the Habs can expect to have their injured players return, those holes still remain unfilled, leaving them with a significant gap at LW in their top-6 and precious little depth should injury strike.

Second, there is the matter of coaching. Martin was capable of forging a deep squad into a very strong 5-on-5 team. After looking over Pittsburgh’s stats during his tenure I’m considerably less convinced that Therrien can do the same, though every indication is that he’s a much stronger coach than Cunneyworth, especially in the area of personnel management.

Barring any upcoming moves by new GM Marc Bergevin, then, it seems unlikely that 2012-2013 will be as promising as 2011-2012 started out, leaving us to lament what might have been, and hoping it will not take several years to build again what the Habs had at the start of last season.

 

(Footnote: For more detail on the subject of analytics, there are already many excellent introductory pieces, including this one by the venerable, terrifying Gabriel Desjardins, a longtime authority on the subject.)

 

Tags: Fancy Stats Featured Habs Jacques Martin Montreal Canadiens Perry Pearn Popular Randy Cunneyworth

  • Layvack

    Very good article, I’m not as woried about Therrien as much as you seemed to be. When he started with the Pens, yes he had Malkin and Crosby, but the team had no wingers for these guys, before Bylsma arrived the Pens top two D-men were injured Gonchar and Whitney, their return coinsided with the arrival of Bylsma. That would make a hell of difference in possesion, as well as games won. Bylsma still looks better mainly because the Pns improved their wingers, so I find it hard to discredit Therrien based on Bylsma’ record or the weakness on the wings and the two top D. It’s easier for a coach to adapte than players and I think the time Ttherrien spent as a news guy will help him a lot, at least, I hope so.

    • MathMan

      The problem is that when Therrien was fired and replaced by Blysma the turnaround was almost instantaneous with largely the same personnel — and no, I don’t buy it was only Gonchar and Whitney, the lack of possession of the Penguins had been an ongoing trend.

      And I mean, it’s a bit of an understatement to say that Blysma looked “better”. The Pens went from “dregs” to “world-beaters”, possession-wise, with the coaching change. Bottom-5 type stats to top-5 type stats. And this has held steady even when Crosby has been out. I doubt that’s just a player thing.

      The closest comparable to the 2007-2008 Pens that year was actually the Habs — weak 5-on-5, weak possession game, strong power play, strong goaltending. That would make Therrien a comparable to Carbonneau, and that is not a good thing.