The Numbers Game: Horse Race in the Northeast

As you sip your morning coffee and peruse the standings after a solid win in Tampa, reveling in the fact that Boston is once again atop the East... wait, what's that you say? Montreal is still ahead? We're tied with Ottawa and Toronto? What fresh hell is this? I know we've got games in hand, but c'mon! Could someone ELSE take the division this year?

Surely there's a way to see into times yet to come and deliver us from this sorcery. Oh great and powerful numbers, I call upon you to look into your numerical wisdom and forecast our divisional future!

PDO - PROBABLY DONE OVERACHIEVING

Fortunately for us there's a statistic out there that can give a glimpse into what teams are over-performing and which have yet to get in gear. I don't want to turn this into a dry lecture on a specific stat, because snooze, but I feel the need to lay some groundwork and put out a few words of caution. So away we go to PDO!

PDO is a pretty simple combination of two fairly basic stats. It's the sum of Shooting Percentage plus Save Percentage. We'll be talking entirely about Even Strength stats for the sake of more predictive results without all that special teams malarky gunking up the data. The idea is that league-wide, each goal that goes in is raising someone's SH% and lowering another's SV%, resulting in the two stats adding up to 100%. Or 1.000 or 1000, depending on how it's depicted and where that pesky decimal point goes according to the person recording the stat. So as a basic rule of thumb, 100% is the norm, the mean, and results above that are outperforming the mean. Over a long period of time and a heaping mountain of data, this statistic has a tendency to regress toward the mean, which is to say things even out around 100% eventually. For a bit more discussion on regression, Ellen Etchingham pretty much ends the need for further writing on it, in crystal clear layspeak, here.

So just as we're getting started, lets throw a grain of salt into the team-wide PDO discussion. Let's say there's a dominant goalie in the league and a shitty goalie in this league. The shitty goalie, lets call him Vesa, by letting in more goals on fewer shots is going to skew the PDO around the league, increasing that of his opponents and decreasing that of his teammates. So while it's true that league wide, 1 is normal, it's a touch reductive. It makes for a good rule of thumb, but please don't mistake it for gospel as there are plenty of factors that can intercede, like Vesa, or like a good possession team keeping the puck on a string and limiting scoring chances. You wouldn't expect Vesa's team to sniff 1.000

We're also going to talk about individual player PDO a bit down below, and with this I want you to put on your thinking cap every time you look at it. We're measuring a player's On-Ice Shooting Percentage - that is, not his own but that of his team when he's on the ice - and On-Ice Save Percentage - likewise, the save record when he's not benched. The idea of regression to the mean of 1.000 is a little too simplistic given the component stats on an individual basis (though if we want to get technical, I suppose it would be correct over a nearly infinite data sample). Not all goalies are created equal, not all shooters are created equal, and not all defensive contributions are equal. Each player's individual stats are going to make up the team PDO, so there's going to be players that perform higher than one, and those that perform lower than one by nature of their talent and roles - it all goes into the team number. Don't make the mistake of automatically assuming above 1 is abnormally high on a player by player basis. it may be "normally" high for them, like Snoop or James Franco. When assessing individual player PDO, take a glance at their past history, at their possession numbers and at their zone deployment. You don't have to make a spreadsheet or do a multi-year deep dive every time you look at a guy's stat, but do consider the influence of other aspects of his game. And especially consider the goalie behind him.

Also, remember that this is a short season. We're not talking about a heaping helping of data. We're talking about a please-sir-may-i-have-some-more spoonful of information from which we're supposed to prognosticate. Weird shit is going to happen. Take our team for instance. Bergeron's a killer defensive forward, it's true, but you probably can't credit him too much for his astronomical 96.2% On-Ice SV any more than you can dog Peverley for his 85.2% (though I will blame Bourque anyway, logic be damned). There will be wild swings, and stuff like this is going to inch closer to normal over time. Again, "normal" for Patrice and Peverley that is, perhaps not league-mean.

AROUND THE DIVISION

PHEW, out of breath. I should probably start some sort of exercise regimen, since I'm just sitting here typing. Anywho, lets take a little spin around the division and see what we can divine.

Below you'll see the latest look at league-wide team PDO as of 2/22. The data's culled from behindthenet, but annoyingly there's nowhere that has team PDO as a sortable stat straight out of the box. NHL Numbers does a nice weekly roundup, but if you feel like futzing with this stuff yourself I'm happy to circulate the speradsheet and its super easy forumla setup as seen here:

First up on our Northeast Division Tour, Toronto, coming in at number 3 on our charts.

TORONTO: 9.7% SH, 93.9% SV, PDO 1036

What can we make of outlier number 1? Did I tip my hand there? Toronto presently has two, count them, two goaltenders with an ESSV over 94%. That's crazy. Let's just stop there and put a little context into the picture. Since the 06/07 season, only three starters have posted above a 94% ESSV for an entire season: Tim Thomas (twice), Brian Elliott, and JS Giguere. Right now this year, there are seven. This is what a small sample size does. It allows outlier performances to take center stage. Over the course of time, barring an exemplary individual talent, the numbers will come down toward league averages. Jonathan Quick is not as bad as he looks now, nor is Craig Anderson as good (we'll get there...). While the weird-ass lockout season of 05-06 did produce three guys at or over the above number, it's highly doubtful we'll see as many as we have now.

Is it possible that Toronto finally sifted through enough trash to find a nugget of gold? As much as it pains me to give Leafs fans a ray of hope, yes it is possible that one of the two guys with tiny NHL samples to draw from is an elite goaltender. We don't have enough historical data on their performance in this league to project how they might continue to fare. I guess they can take some sort of positive out of the fact that they have the same coach Giguere posted his good year under, though surrounding mediocre years don't lend a whole lot of support to the Carlyle system theory. Under whose systems, by the way, they are now facing more shots from roughly the same range.

So, it's in the realm of possibility that a stellar netminder has emerged from Toronto. Bully for them, we "emerged" their last one for them. So lets poke some holes in the balloon. Remember how we talked last time about Fenwick Close, and how it's possession picture is predictive of production? Well here the Leafs don't fare very well. They rank in the bottom 5 in terms of possession, which indicates that their opponents are heavily outshooting them, lending a little credence to the assumption that luck is pretty heavily at play in Toronto's performance to date. When the game's tight, they're not driving the play their way and they're being buoyed by their shooting accuracy and stellar save count. Since they don't have the puck much, their third-in-league goaltending could bail them out as it has thus far, or their fourth-in-league shooting percentage could stay where it is. But counting on both?

To quickly further address the shooting side, Phil Kessel will not - repeat, will not - continue to shoot 6% below his career average. On the other hand, Tyler Bozak and Matt Frattin are riding absurd percentages. We are not witnessing the emergence of a PPG player in Frattin, given that he probably won't maintain 40% shooting and doesn't have an NHLe history to back up. The overall team shooting isn't wildly out of hand league-wide, but it is a good percent and a half over average and features several uncharacteristic performances, particularly on the blueline. Moreover, it's uncharacteristic for the team, as they are vastly outstripping the past few years' accuracy.

All to say, Toronto's riding high on luck and should be hoping that they've banked enough to get them into the playoffs in a shortened season because unless Scrivens does prove to be Thomas' offspring, they're in for a drop.

MONTREAL: 8.8% SH, 93.2% SV, PDO 1020

Nobody here wants to hear this, so skip to the next if you like to read about terrible things happening to our opponents, but Montreal isn't really due for that much of a course correction. Carey Price's 94.1 ESSV is a bit high, but he has posted 93 in the past and usually hovers around 92. So perhaps you can expect his side of the equation to come down a tad, but the shooting numbers are pretty normal and middle-of-pack.

Like Toronto, they have a few guys who are shooting the lights out, but no one on the absurd end of the scale. Some of the big guns and defensemen, especially Markov if he doesn't hit the IR first, should come down a touch from their high teens numbers over time, but their bottom six are underperforming pretty severely (sound familiar?) so the overall number should remain fairly constant team wide. They are right now shooting in line with last year, though prior years were considerably worse. Somewhere beween, around 8%, is likely.

Montreal had a lot go wrong last year while maintaining pretty decent underlying possession stats. They're fairing well there again this year, while getting the goaltending and goal support they lacked before. Annoyingly, we'll probably be neck and neck for much of the year and so will need to make the most of those games in hand. Head to heads will be critical.

OTTAWA: 5.9% SH, 94.2% SV, PDO 1001

Losing Erik Karlsson is a Norris sized blow to the blue line (...gag) that's sure to have some impact on their play, but we're not here to talk about that. This is probably the most readily legible case study in the whole bunch for the purposes of examining PDO. What we've got here is an unsustainable goaltending performance, coupled with an unsustainably low shooting percentage. Anderson is posting a 95.3% ESSV, which would be a record over a full season, and the team SV% is approaching Tim Thomas in 2010/11 levels. (To sidetrack for a moment, Tim was not bested by Elliott's fewer games the following year at ES. So good job on the PK, St Louis. Eat it, Elliott). Don't get me wrong, Anderson's had some nice outings in the past, but this is an outlier to end all outliers and I cannot find any kind of historical support that says he can maintain close to this rate over the long haul.

On the other end of the ice, only Edmonton is hitting the back of the net less. On the whole team, only Jason Spezza is registering a good On-Ice SH%. While the SV% should come crashing back into the ionosphere, they'll get a little dead cat bounce from the shooting. Over the past three years, Ottawa's been pretty lacking in SH%, having only three guys last year with an On-Ice above 10%, so I don't expect the two sides of the story to cancel each other out. The goaltending will drop sharply, while the shooting will rise modestly, based on historical data. It will probably be enough to keep them out of the playoff picture if the Conference race stays tight.

BUFFALO: 9.1% SH, 90.9% SV, PDO 1001

Buffalo appears to be what Buffalo is, more or less. They are getting a little bit lucky with shooting, largely driven by the Thomas Vanek Explosion, keeping them a full percent above their three year average. But any crash on this side will likely be offset by Ryan Miller, ES picture of consistency, who with a large workload will drag their SV% toward his typical 92%.

However, we can't extrapolate quite as much from the numbers with Buffalo as we can with other teams, as we don't have any kind of background as to how players will be deployed under a new coach. Suppose he does something crazy logical like employ Vanek on an exploitation line or actually uses Ott to win the faceoffs they're getting killed in. Starting with the puck would make a huge difference for this team, who are getting Corsi crucified because of this deficiency. But Lindy Ruff was a coaching deity (and not a guy propped up by Hasek, no sir), so it's best to assume nothing but decline here on out since no one will ever compare to his genius.

BOSTON: 8.8% SH, 90.4% SV, PDO 992

Last but not least, we come to the underachiever. Sporting the lowest PDO in the division, and bottom 10 league-wide, Boston is the one major candidate for improvement. Presently their SV numbers are skewed by a couple of weak goaltending outings. As discussed ad nauseum, we're winning games by thin margins and, while not losing much, our goal differential isn't too stellar. The two Buffalo games alone are enough to drop our overall SV%. For those aghast at the number above, take solace in the fact that Rask is a snuggly middle-of-the-road 91.6%, who independent of Khudobin is enough to raise our PDO just above 1000. With more games recorded, barring another visit from cyborg assassin Tomas Vanek, Rask will raise the ESSV toward respectable.

Shooting is a bit difficult to suss out based on past years' data. 09/10 and 11/12 are such wild outliers that trying to draw a conclusion over recent seasons is pointless - just like 09/10. Assuming that the Cup year's league-average shooting is the norm, we're pretty much where we should be, in spite of Brad Marchand's ridiculous 45% pace. Meanwhile, that's offset to some degree by Bergeron's meager 6.7% rate and Tyler's 7%. Their collective On-Ice is pretty much where it should be. And as Phunwin astutely points out, Brad's barely getting off a shot per game. Should he pick up the pace in overall shot count, it will compensate for - if contribute to - the coming drop in his accuracy without affecting his output.

We are "fortunate" that we have several players whose PDO is abnormally low. Peverley and Kelly in particular, though saddled with babysitting duty, are more than due for some bounces to go their way, and for Rask's output behind them to head upward. But don't expect Kelly to replicate last year - that was above his normal by a ways.

Taking into account our third-in-league Fenwick Close, which has helped prop us up in the same way Toronto's luck has, we're significantly underperforming and with good netminding should actually expect our PDO to exceed 100% this year. We own the puck, we're getting the goaltending most of the time, we just need to see an uptick from a few regulars and bam, zoom, straight to the top!

OVERALL

Over the long run, it looks like PDO is on our side. However, as we've examined the division's performance to date keep in mind that emphasis on "the long run," because in a shortened season there's not as much runway for a low PDO to takeoff, and not as much time for the inevitable decline to befall the fortunate. Which isn't to say that I've wasted your morning or lunch hour with a bunch of bullshit mumbo jumbo, just that I've wasted your morning with hypotheticals based on likely outcomes. Outcomes such as a decline from Toronto and Ottawa to get them off our tail, Buffalol continuing to be mired in mediocrity, and Montreal posing the only real threat to our supremacy. Isn't that just the way rivalries should be?*

*no, screw them.