Predicting the Unpredictable, Expected PDO: Atlantic Division

With the rise of the influence of the analytics community over the landscape of hockey, many new statistics have come to the forefront of forward-thinking analysis. One of those stats, known as PDO, was created with the primary goal of quantifying how lucky or unlucky a given team or player has been.

The statistic is largely used only with 5-on-5 play. It is comprised of on-ice save percentage and on-ice shooting percentage. When the two are added together, the prevailing thought is that over time, most teams will trend toward the mean value of 100. However, it is rather rare that teams find themselves with flat 100 PDOs.

Intuitively, it makes sense that using the same number and applying it to all 30 teams as a sort of “expected” value may not hold up very well. It isn’t luck that Henrik Lundqvist and Cam Talbot provided the Rangers with a better on-ice save percentage than the one earned for San Jose last year by Antti Niemi and Alex Stalock, for example. Lundqvist and Talbot are just better goalies. Therefore, it is to be expected that teams with demonstrably better goalies than average would have a leg up in what their PDO number would look like in a given year.

The same principle applies for shooters. While it’s a known fact that much of what goes into shooting percentage over shorter sample sizes is heavily based on random variance and puck luck, it’s also true that there is some level of correlation between shooting percentage and player quality. Steven Stamkos doesn’t have a career shooting percentage of 17.2% (compared to league average at ~8.5%) by accident; he’s just an excellent shooter. Take a look at New Jersey’s forward group heading into next year, then have a gander at Tampa Bay’s. Certainly, you wouldn’t want to hold them as teams to the same standard in scoring efficiency that would define them as “lucky” or “unlucky.”

Because of this, I’ve decided to make an attempt at my own version of an Expected PDO. Using a three year sample, I’ve compiled even strength data on the 20 skaters for each team most likely to play significant minutes over the course of the season, as well as each team’s starting and backup goaltenders.

Be sure not to take what I’m doing out of proportion. I’m not trying to predict PDO because that has proven to be practically impossible. I’m fully expecting pretty much all of these numbers to end up being wrong by the end of next season, just as most teams check in above or below the current baseline PDO of 100. I’m just curious to see if some minor alterations to account for player movement combined with past team performance over a sizable sample can do a better job of predicting a team’s PDO than assuming at the beginning of the year that each team will be at about 100, and anything considerably above or below that is almost entirely luck.

Pension Plan Puppets did something similar (just with the Maple Leafs only) in the middle of the 2013-2014 season that is fantastic and entirely worth a read.

With that being said, doing all 30 teams in one post would be a behemoth task and impossible for you all to read, so I’ve decided to split it up by divisions. I’ve already posted the Metropolitan Division’s breakdown on the 23rd, today I’m moving on to the Atlantic Division, followed by the Central Division on the 25th, and concluding with the Pacific Division on the 26th. By the end of it all, we should have some pretty interesting numbers crunched out to critique and speculate about heading into next season.

So without further ado, here’s the data for the Atlantic Division:

1. Montreal Canadiens: 101.85

The Montreal Canadiens are the team with the highest Expected PDO in the entire league according to my model. This is the Carey Price effect in full force. The reigning Hart Trophy winner and Dustin Tokarski project to provide the bleu, blanc, et rouge with a 93.3% clip at saving even strength shots.

There are also a number of forwards on this roster that seem to possess the ability to drive up shooting percentage. Over the sample used in this project, the following players own an even-strength shooting percentage higher than 10%: Max Pacioretty, David Desharnais, Zack Kassian, Alex Galchenyuk, and Dale Weise. High caliber offensive defensemen such as P.K. Subban and Andrei Markov also provide the Canadiens with an above average defensive unit with regards to scoring efficiency.

The end result of that is a strong expected shooting percentage of 8.55%. The Atlantic division has by far the highest average Expected PDO of any of the four divisions in the league. That’s something you’ll notice as this post progresses. You may be surprised to learn that over the last three years, 18 of the league’s 30 teams have PDOs above 100. My theory on this is that there is likely far more significant turnover in goaltenders league-wide than there is at skater positions. Because of this, the weaker goalies that give up a lot of goals are out of the league more quickly and upgraded upon, thus driving up teams’ save percentages but not affecting their own shooting percentages.

2. Boston Bruins: 101.63

Similar to Price’s influence in Montreal, Tuukka Rask’s reliably consistent excellence in goaltending gives the Bruins a massive leg up in what their PDO should look like at the end of any given year. It’s tough to project how Malcolm Subban will perform as Boston’s backup goaltender given the small sample size of his play last year, but my model expects a league-best 93.5% even strength save percentage out of the Bruins. I would bet on the lower side of that if I could, but it’s hard to imagine a team with Rask as its starting goalie coming in below 93.0%.

With regards to skaters, the offseason moves the Bruins have made seem to suggest that they’ve brought in some high percentage shooters. Matt Beleskey, Jimmy Hayes, and Brett Connolly are each above 10% at even strength over the last three years. However, Connolly’s sample size in particular is far too small to expect that he’ll be able to maintain that. It’s also possible that Beleskey and Hayes regress, but either way this forward group is full of players who are efficient scorers. The end result is an 8.13% mark in the expected shooting percentage department.

3. Ottawa Senators: 101.06

The number my model produced for Ottawa is almost identical to the 101.0 PDO they were good for last season. I do have some reservations when it comes to selling you on this number, because I do expect it to be slightly lower in actuality.

Craig Anderson is an excellent goaltender at even strength, but Ottawa’s expected save percentage of 93.3 is driven up from Anderson’s number by Andrew Hammond’s performance last season. I don’t at all expect him to be able to replicate what he did last season, even if he’s still a serviceable backup regardless.

With their skaters, the number may be a little bit wonky due to the small sample sizes from the heroic performances Mark Stone and Mike Hoffman last season. However, because these are small sample sizes, their influence on the overall team number is inherently limited. So ideally it will be balanced out a little bit. This should be a strong offensive team next year. It’s likely that Erik Karlsson is capable of having a positive impact on his team’s on-ice shooting percentage just by nature of how excellent of a player he is.

The Senators check in with an expected shooting percentage of 7.76%, and I’m very interested to see how they perform in actuality in comparison to that number.

4. Toronto Maple Leafs: 101.01

And here we are, finally arriving at the division’s most intriguing case in advanced statistics. The Maple Leafs’ expected number definitely took a hit when they dealt star scoring winger Phil Kessel to the Pittsburgh Penguins.

The solid but unspectacular duo of James Reimer and Jonathan Bernier between the pipes delivers a solid but unspectacular expected save percentage of 92.3%. I see no good reason whatsoever that they won’t be able to deliver that at the very least.

They made several keen moves in the offseason that could serve to improve their possession and percentage numbers marginally, but nobody is fooling themselves into thinking this team is going to be anything resembling good next season. The rebuild is what it is, and in my mind they’re doing a good job following through on it.

The expected shooting percentage number of 8.71% is a bit puzzling. I figured heading in that with Kessel gone that number would suffer a lot more. As it turns out, Tyler Bozak, Nazem Kadri, and Joffrey Lupul all exert major positive influences on Toronto’s shooting percentage numbers. The newly acquired Shawn Matthias also brings some serious scoring efficiency with his 37 goals on 309 even strength shots over the past three years. As always, it will be highly intriguing to watch and find out how the Maple Leafs perform this season.

5. Tampa Bay Lightning: 100.84

As an already demonstrably elite team when it comes to puck possession, the reigning Eastern Conference champions look to be in store for another fantastic season should this hold true. Ben Bishop and rising star backup Andrei Vasilevskiy are expected to provide their team with a solid 92.7% mark in the net.

You may expect Tampa Bay to have one of the best expected shooting percentages in the league, but that isn’t quite the case. They are above average, but their 8.14% number is slightly surprisingly low. Erik Condra is a very solid acquisition for their forward depth. Their big four forwards of Steven Stamkos, Tyler Johnson, Ondrej Palat, and Nikita Kucherov are all far above average as a group when it comes to shooting percentage, but their defensive unit as a whole is one of the least efficient in the league at turning their shots into goals. That may have a lot to do with why their expected shooting percentage isn’t quite at the top of the league.

6. Detroit Red Wings: 100.38

The Red Wings are now entering the post-Babcock era as this season approaches. His successor Jeff Blashill seems to be a similarly minded guy, and I’m excited to see what slight differences in his system reveal themselves in the early going in October.

I think my projection may be overestimating what the tandem of Petr Mrazek and Jimmy Howard will be capable of next season. I fully expect Mrazek to develop into a star goalie, but the 92.8 mark (even with the New York Rangers) my model came up with may be a little overzealous, especially given Howard’s struggles of late.

Like the Lightning, the Red Wings are another team that at face value, one would expect to have a high expected shooting percentage that does not. Their number is a meager 7.58%, and could be explained by a general overestimation of the percentage driving abilities of Henrik Zetterberg, who has only converted on 24 of his 363 even strength shots over the last three season. If he comes back from that, the Red Wings should be well positioned to outperform this number.

7. Florida Panthers: 100.2

The Florida Panthers are a team on the rise with a reasonably strong possession game and an impressive young core. With solid veterans like Roberto Luongo and Al Montoya holding down their roles as starting and backup goalies, there’s a lot to like about the Panthers heading into next season. The expected save percentage for that duo checks in at a respectable 92.4%.

Florida’s expected shooting percentage, on the other hand, is 7.8%. That’s not a great mark, but it’s not really bad either. That evaluation matches up pretty well with what I think we can expect out of Florida’s forwards this season. The continued development of Aleksander Barkov, Nick Bjugstad, and Jonathan Huberdeau combined with the veteran leadership and production of Jaromir Jagr should help serve to offset their lack of proven depth throughout the lineup. I expect the end result to be a pretty good group of players that could compete for a playoff spot as long as their PDO doesn’t fall too far below this number.

8. Buffalo Sabres: 99.48

It should come as no surprise that the lone sub-100 PDO team in the Atlantic Division is the hapless Buffalo Sabres. The main reason that their number is this respectable is the acquisitions of high percentage shooters such as Evander Kane and Ryan O’Reilly.

The inexperience of their defensive unit may also have an inflationary effect on their expected shooting percentage of 7.58%. The sample size from those players (Ristolainen, Colaiacovo, etc.) isn’t quite enough at this point to properly balance out the influence their forwards have (which obviously would be greater regardless).

The expected save percentage from recent acquisitions Robin Lehner and Chad Johnson comes in at 91.9%. This isn’t atrocious, but you’ll note that it’s certainly not good. The Sabres should be vastly improved with O’Reilly, Kane, and Jack Eichel in the fold, but expecting a playoff berth at this stage of their rebuild would be rather optimistic.