Dustin Byfuglien was doing things to make fans of the Winnipeg Jets worry before he’d even stepped on to the ice in the team’s new jersey. In August, the defenseman was arrested under suspicion of impaired boating; perhaps worse from a hockey perspective was the fact that his weight was reported at 286lbs – well above his playing weight.
It wouldn’t be a welcome break for any hockey player, but given Byfuglien’s late-season slide it was particularly concerning. In the first 41 games of 2010-11, Byfuglien’s offense was off the charts, as he recorded 15 goals and 25 assists for a total of 40 points, along with a plus-6 rating. Over the final 40 games, he tallied just five goals, eight assists, and 13 points, and went minus-8 to boot.
Now, Byfuglien is minus-6 just eight games into the season. Is he in real trouble?
Before we compare the two halves of Byfuglien’s 2010-11, let’s start by looking at his underlying numbers in the early going.
First off, Byfuglien’s playing a lot against the best opposition. Here are his most frequent opponents for every game this season (courtesy of timeonice.com):
- October 24 – Brad Richards’ line
- October 22 – Eric Staal’s line
- October 20 – Jason Spezza’s line
- October 19 – Phil Kessel’s line
- October 17 – Jordan Staal’s line
- October 15 – Shane Doan’s line
- October 13 – Jonathan Toews’ line
- October 9 – Andrei Kostitsyn’s line
It’s too early for the Quality of Competition numbers at Behind the Net to be useful, which is why I went through the games manually, but I think the pattern is clear: Dustin Byfuglien is spending a lot of time on the ice against the opposition’s best. He and Tobias Enstrom are also getting more own-zone work than one would typically expect of a high-end offensive defenseman – among regulars, only Johnny Oduya is used more in the defensive end of the ice.
I go through all this for context – it’s important to establish what situations a player is used in, and Byfuglien’s been handed tough minutes. How has he performed? Let’s look at some data from 5-on-5 situations:
There are a bunch of acronyms in the chart above, but really it’s pretty simple. After listing names and games played, we’re looking at an average 60 minutes of five-on-five ice-time in two different ways: goals for and against while the player is on the ice, and shots for and against while the player is on the ice.
The goal data is straight-forward, and Byfuglien’s numbers are awful. The team scores less goals with him on the ice than with any of the others, and allows more too!
But the shot numbers paint a different picture. The Jets average 30.1 shots for every 60 minutes of 5-on-5 time with Byfuglien on the ice; that’s the best number in this group. His shots against number is also quite low, and when we combine those two figures, we find something we wouldn’t have expected from the goal data: the Jets outshoot their opposition more with Byfuglien than with any other defender except Tobias Enstrom.
How is Byfuglien a minus-6 if the shot numbers are so good? It comes down to the percentages. With Byfuglien on the ice, the Jets have scored on about one in 20 shots. The opposition, on the other hand, is scoring three times for every 20 shots. This isn’t because Byfuglien’s opponents are taking shots that are three times as good as Byfuglien and his line-mates. Rather, it’s just a fluke result, something that will even out as he plays more games.
PDO is a measure of percentages – it combines on-ice save percentage with on-ice shooting percentage – and it’s a good way to see which players have been overly impacted by luck (on-ice percentages aren’t entirely luck – quality of teammates, opponents, and other factors play in – but a big chunk of the number is random). Last year, the player with the league’s worst PDO and at least 40 GP was Ottawa’s Filip Kuba – with an on-ice shooting percentage of 6.54 and an on-ice save percentage of 88.3, he had a total PDO of 94.9 (100.0, obviously, is league average). Byfuglien’s PDO so far this year is 90.0 – so bad that it can’t possibly continue.
The percentages also give us a window into what happened last season. Let’s split the year into two halves, along the lines mentioned in this piece’s introduction: the first 41 games together, the last 40 games together. Here is on-ice shot, goal, and percentage data at even-strength for Byfuglien last season:
|Time Frame||Shots +||Shots –||Shots +/-||Goals +||Goals –||Goals +/-||On-Ice SH%||On-Ice SV%||PDO|
|First half, 2010-11||426||381||45||40||26||14||9.39||93.2||102.59|
|Second half, 2010-11||482||369||113||34||34||0||7.05||90.8||97.85|
The Thrashers actually had better shot data with Byfuglien on the ice in the second half of 2010-11than in the first half; yet, his plus/minus went down. Did his diving plus/minus reflect a drop-off in play? I would suggest the answer is no.
Byfuglien’s offensive numbers are another matter. Once again, we’ll split his year in half, and look specifically at even-strength and power play scoring, as well as shooting percentage and shot totals.
|First half, 2010-11||10||13||5||12||162||9.3|
|Second half, 2010-11||2||4||3||4||185||2.7|
Byfuglien’s shooting percentage is the obvious place to start. His torrid beginning to the season probably set unrealistic expectations for him – Byfuglien’s career shooting percentage prior to joining Atlanta was just 8.6%. 9.3% doesn’t sound like a big jump on 8.6%, and it isn’t normally, but Byfuglien’s circumstances were unique because he changed position from forward to defense. Defensemen never put up the same kind of shooting percentage numbers that forwards do; it was unreasonable to expect a typically low-percentage forward like Byfuglien to suddenly turn into a high-percentage defenseman.
Byfuglien’s shooting percentage came crashing down to Earth in the second half. It’s a much better idea to bet on him finishing near last season’s overall shooting percentage number this year rather than the overly high total from the first half the season or the overly low total from the second half.
The assist data, particularly at even-strength, is more difficult to explain. Byfuglien was on the ice for 40 even-strength goals in the first half, assisting on 13 (32.5%) of them. In the second half, he was on the ice for 34 even-strength goals, but managed just four assists (11.8%). He was almost three times as likely to record an assist on a goal in the first half as he was in the second.
Scott Reynolds of Copper & Blue came up with an interesting statistic a few years back to explain that kind of variance in assist totals – Individual Point Percentage. What Reynolds did was look at defenseman with a minimum of 50 games played, and figure out what percentage of goals they recorded a point on. He found that the average defenseman recorded a point (either a goal or an assist) on 28.6% of the goals he was on the ice for at even-strength.
In the first half of 2010-11, Byfuglien’s IPP was 57.5%, a total higher than the four-year average of any of the players Reynolds studies – and thus a total unlikely to continue. In the second half, Byfuglien’s IPP dropped to 17.6% – a total lower than the four-year average of every defenseman except for Greg Zanon (and thus a total unlikely to continue).
The important thing to note about all of these numbers is which ones vary wildly over the course of a season, and which ones remain steady. The statistics that remain steady – shot-based metrics, for the most part – were good for Byfuglien in both the first and second halves of 2010-11. By those measures, he was a good defenseman throughout. The statistics that vary wildly – on-ice percentages, PDO, IPP – were unbelievably good in the first half and incredibly bad in the second half.
In short: it’s very unlikely that Byfuglien’s second-half collapse last season was because he suddenly forgot how to play the game. It’s far more likely that his performance in 2011-12 is less eratic, falling somewhere between the extremes of last year.
Going back to the introduction, we asked whether Byfuglien’s poor start and late-season collapse last year meant that he is in real trouble as a high-end defenseman. Based on the data, the answer to that question is ‘no.’