Jeff Marek, the co-host of sportsnet.ca’s Marek vs. Wyshynski program, pointed out an uncomfortable fact about Ryan Nugent-Hopkins on Tuesday. In the course of a conversation with Greg Wyshynski about Nail Yakupov being the first of the Edmonton Oilers’ young players to be scratched, Marek asked, “why is it not Ryan Nugent-Hopkins, who wears the green jacket at nine under in the NHL?”
The problem with Marek’s question is that plus/minus, while a great concept, is a terrible statistic. Many of its flaws are well known; the most important flaw, the randomness of shooting and save percentage, are not.
The major advanced statistics websites tend to present five-on-five (5v5) rather than even-strength (ES) data. Four-on-four (4v4) and three-on-three (3v3) happen infrequently, and the goal is to make apples-to-apples comparisons, so it’s easiest to only present 5v5 data. Most of plus/minus is 5v5 play, so we’ll work with that data.
Through the Oilers’ game against the Washington Capitals Monday, Nugent-Hopkins had been on the ice for zero 5v5 goals for and five 5v5 goals against with the goalies in the nets. Sounds awful. It’s not minus-9, but a star player on pace to go minus-82 over the course of a full season seems problematic. So was it reasonable to point to his plus/minus as a sign that he wasn’t playing well?
Well, at the time of Marek’s comments, Nugent-Hopkins had been on the ice for 54 5v5 shots attempts for and 56 5v5 shot attempts against. Viewed that way, Nugent-Hopkins’ play looks much better. If you only consider shots on goal, things look even better: the Oilers took 31 5v5 shots and allowed 21 5v5 shots with Nugent-Hopkins on the ice and both goalies in the nets through his first four games.
So how do you go minus-5 at 5v5 while outshooting the opposition 31-21 over four games? Percentages. Namely, shooting percentage and save percentage. The Oilers had a 0% shooting percentage and a .762 save percentage at 5v5 with the goalies and Nugent-Hopkins on the ice over his first four games of the season. It’s fair to say the Oilers getting outscored 5-0 with Nugent-Hopkins on the ice at 5v5 with the goalies in through four games had nothing to do with the volume of shots and everything to do with pucks going in.
“OK,” you might say, “so Nugent-Hopkins has a terrible plus/minus because the Oilers can’t get a save with him on the ice and can’t score a goal. He must be generating nothing and making defensive mistakes all over the place.” It’s possible that this is the case; you can’t disprove it with the numbers that are currently available. What you can do, though, is show that on-ice save percentage seems to be something over which a player doesn’t have much of control.
We can illustrate this by looking at players who had a really high or really low on-ice save percentage in the first half of a season and compare it to their on-ice save percentage in the second half of the season. We’ll use the 2011-12 season, which was the last complete season the NHL played, to illustrate the point. Here are the skaters with the best 5v5 on-ice save percentages in the first half of the season (minimum 200 shots in each of the first and second halves):
That’s an impressive average performance from that group — the league average save percentage at 5v5 is about .921. A goalie who posted a .962 5v5 save percentage over the course of an entire season would have an argument that his season was more valuable than any in history. We’ll call this bunch the best group. What about the worst performers (worst group)?
Horrific. That group averaged an .895 save percentage at 5v5, a number that would cost a goalie his job. The difference in terms of goals allowed is amazing and should be noted. On average, the players in the worst group were on the ice for nearly 24 more goals each in the first half. That will have a significant impact on a player’s plus/minus.
Here’s where things get interesting. Hockey analytics people tend to think that a skill is something that persists. For example, the players who get the most points in a season tend to score more points the following season than the players who get the fewest points. We can be comfortable that accumulating points is a skill. If posting a high or low on-ice save percentage is a skill, we’d expect to see the players who did best at it in one half of a season be better than the players who did worst at it. How did the best group do in the second half of the 2011-12 season?
That looks awfully close to that league average of about .921 I mentioned earlier. What about the worst group? How did they make out?
They actually posted a better on-ice save percentage in the second half of 2011-12 than the best group. If on-ice save percentage is something significantly controlled by the players on the ice, for some reason the players who were best and the players who were worst in the first half suddenly became awfully average in the second half.
An alternative explanation for this phenomenon is that players at the NHL level aren’t exercising a lot of control over whether a puck fired at the net goes in. To the extent that they are exercising control, it’s less important than the randomness of hockey. Sometimes players go through stretches where the opposition makes their shots; sometimes they go through stretches where they don’t. As it turns out, operating on this assumption gives you a good starting point when discussing players who have a really bad or really good on-ice save percentage. If you act on the assumption that it’s fortune, whether good or bad, you’ll rarely be wrong.
As far as Nugent-Hopkins goes, he’s stemmed the bleeding in his last two games. The Oilers’ 5v5 save percentage with him on the ice has been .950, and they’ve shot 12.5%, leaving him even despite being badly outshot in two games against the Pittsburgh Penguins and New York Islanders. He won’t run up against Sidney Crosby and John Tavares every night. And, as is set out above, it’s likely that the save percentage behind him and, by extension, his plus/minus, will take care of itself as time passes.