I’ve posted it many times before but for the newly-initiated, here’s a chart that shows just how predictive Fenwick can be. For this season, there were two teams with at least a 55% Fenwick Close%: the Los Angeles Kings (last year’s Cup winner and 2013 Western Conference Finalist) and the Chicago Blackhawks (the West’s representative in the Stanley Cup Finals this year). Down the road, there will be other ways in which we will be able to probabilistically predict who should win a hockey game, but there are no such tools available at present time.
The idea behind using possession as a predictive tool is two-fold:
- Evidence shows that teams are unable to drive the number of scoring chances, rather the ratio of scoring chances/all shots is fairly consistent among teams. The reasons for this are shown in Part One of my look at how Pittsburgh faltered. This is why possession is so highly valued; if you can’t greatly alter the ratio of scoring chances, then by having more chances overall, by extension, you will have more scoring chances.
- Also in Part One, I discussed how wildly varied a player’s shooting percentage can be from year to year. The example I give is that over an 82 game season, the difference between Alex Ovechkin’s highest and lowest-shooting percentage seasons can mean a 25 goal swing. The best way to mitigate this variance in shooting percentage is with shot volume (which is what Corsi and Fenwick measure).
Earlier (Part One), I looked at how Pittsburgh is attempting to evolve from shot volume and use their wealth of superstars, namely Sidney Crosby and Evgeni Malkin. The idea is that with such highly-skilled players, a coach can design line combinations/offensive systems around this notion and management can acquire players whose talents can be maximized by being put with Crosby and Malkin. This line of thinking is what led to the acquisition of James Neal.
The Playoffs
Pittsburgh finished first in the East this year, a full nine points ahead of their next-closest challenger (a pretty significant edge in a 48 game season). They finished first in the NHL in goals scored despite a 49.9% FenwickFor%, good for 15th in the NHL and only slightly ahead of the Minnesota Wild. What this means is that at 5 on 5, over the course of the season, they were out-shot attempted by their opponents (not including blocked shots). Some of this of course is due to score effects (how the score of a game can affect a team’s strategy), which is why Pittsburgh was a plus FenwickClose% team (when the score is within a goal in period 1 or 2, or tied in period 3). They managed to finish the top scoring team in the regular season because of their power-play (2nd overall) and their team 5 on 5 shooting percentage, which was T-4th in the NHL.
All this is to say that through the regular season, Pittsburgh’s attempt to drive scoring chances instead of shot attempts appeared to be working.
Pittsburgh vs New York Islanders
People gave little hope to the Islanders (I predicted Pittsburgh in 5 games). This series ended up going six. It’s not a surprise that Pittsburgh won. How they went about the series win, however, was a warning flag.
- Pittsburgh was out-attempted by the Islanders 312-261 at 5 on 5 over the six games. On average, the Islanders had 8.5 more shot attempts at 5 on 5 than Pittsburgh did per game.
- The Islanders managed 15 even-strength goals against Pittsburgh (despite being shut-out in two of the six games). The Penguins managed 18 even-strength goals on the Islanders. So at even-strength, the Penguins were much more efficient on their shot attempts (6.9% of Corsi events turned into goals) than the Islanders were (4.81%). A new team strategy working? Perhaps.
- Pittsburgh was exceptional on the power-play. The Penguins scored seven of their 25 goals in this series with the man advantage, reeling off an outstanding 33.3% efficiency (7 for 21 power-play opportunities). Scoring at that rate, with 21 total power-plays, meant Pittsburgh was pretty close to up 1-0 in every game before it even started.
Goaltending was an issue for the Islanders. Despite their talent up front, Evgeni Nabokov couldn’t stop a proverbial beach ball, finishing the series with an .842 overall save percentage and an .863 even-strength save percentage. It was as if Pittsburgh was being too efficient (Nabokov did have an .893 SV% against Pittsburgh in the regular season in five games; still bad, but nowhere near that putrid .842).
Pittsburgh vs. Ottawa Senators
This round would seemingly provide a better test for the Penguins. Ottawa is a much better team defensively than the Islanders and goaltender Craig Anderson was being bandied about as a possible Vezina Trophy nominee as the NHL’s top goaltender (not by me, however). They didn’t pack the offensive punch that the Islanders did, but surely if anyone could slow down the juggernaut that is the Pittsburgh Penguins, it would be one of the best goaltenders and defensive teams in the NHL, right? Well…
- Pittsburgh scored 22 goals in five games, an even more prolific rate than they did against the Islanders.
- Pittsburgh kept their power-play rolling, scoring six goals on 25 chances. That 24% success rate was lower than the first round, but right in line with the regular season rate of 24.7%.
- Again, Pittsburgh was out-attempted over the course of the series 246-228 by their opponent. However, as was the case earlier, Pittsburgh was again more efficient at even strength, scoring 12 goals on 228 Corsi events (5.26%) against Ottawa’s six goals on 246 Corsi events (2.44%).
After their first two series (11 games), Pittsburgh held a significant 13-4 scoring edge on the power-play against their opponents and were significantly more efficient with their 5 on 5 Corsi events; their 6.14% efficiency dwarfed the Islanders/Senators combined efficiency of 3.76%.
Pittsburgh looked to be rolling and their attempt to buck the “teams can’t drive scoring chances” line of thought was one series win away from proving itself worthy.
Pittsburgh vs Boston Bruins
It had been well-established that these two teams seem almost destined to meet in the Conference Finals. And what a dream match-up. The Offensive Machine vs. The Defensive Stalwarts.
Except this wasn’t the Islanders, who had a goalie in name-only. This wasn’t Ottawa either, who didn’t have offensive weapons in abundance like Boston did. This was a complete team they were facing.
Two trends emerged that were radically different from the first two rounds:
- Pittsburgh out-attempted their opponents at 5 on 5 over the course of the series 182-164. Not a significant advantage, but certainly different from what they had shown earlier in the playoffs.
- Pittsburgh put up a goose egg on the power-play. The team that had scored 13 PP goals in 11 games went 0-15 in four games against Boston.
Pittsburgh scored their only two goals of the series at even-strength. Boston scored 10 of their 12 goals in the series at even-strength. This gave Pittsburgh a Corsi efficiency of just 1.10%, while Boston’s was 6.10%.
What Does All This Tell Us
We need a lot more information to accurately analyze this. Things like scoring chances and shot distance have a great effect on shooting efficiency (which is what Pittsburgh is aiming for). Without a centralized database, it’s tough to track these things (there is a shot database, but it’s not updated for postseason play yet).
Hopefully, someone with greater technical knowledge than I will figure out a way to centralize this data. Until then, this is my educated guess:
Any model that relies on result (goals) instead of process (shot attempts) will be subject to higher variance, regardless of the skill level on the ice. There is little evidence to suggest that the skill level of an individual player can influence more than a handful of outcomes in a given season. By extension, yes, a few highly-skilled players means more influence, which means more wins. But eventually, the variance has to catch up somewhere.
Remember, variance can be good and bad. Maybe a team has a stretch of a few games where they score four or five goals instead of two or three. That also means there will be also be stretches where they score one or none instead of two or three.
To be sure, some credit goes to Boston. Their penalty kill was excellent and Tuukka Rask was unbelievable in net. On the flipside, there were several instances that led directly to goals where Pittsburgh simply seemed disinterested in playing hard defensively, as I pointed out in Part One. Boston was willing to do anything necessary to win hockey games, Pittsburgh was not.
All that said, there had to be a significant luck element here. That luck element fell on the wrong side of four games for Pittsburgh. That is why perhaps their model could be flawed; it’s totally dependent on the result instead of the process.
It’s not like it should come as a huge shock that they could have simply been on the wrong side of their self-created variance. This season alone we saw this happen (to a somewhat lesser degree): From January 23 – January 29, a span of four games, Pittsburgh only scored a total of six goals against some rather shaky teams (among them, Toronto, Winnipeg and the Islanders). From March 17 – April 5, a span of ten games, they scored two or less goals in eight of them.
Coaches always preach process over results. Pittsburgh was (is) concerned with results over process. This attitude certainly bore out in their quality of play in the Boston series.
This is not to say there isn’t something to trying to create more scoring chances, but an entire organizational philosophy built around a model that is intrinsically more variant than what just about every other team employs? There’s a reason why the two Stanley Cup finalists were #2 and #4 in Fenwick Close% this season.
