Considerably less surprisingly, I disagree with all of them.

To be fair, Ovechkin has had a good season and there really aren’t any forwards that have stepped up and separated themselves from the pack. Still, giving Ovechkin the MVP would be a travesty.

Mainstream analysts see that Ovechkin’s 47 goals leads the league by 7, he has a +/- of +11 -- which looks positively stellar compared to the disastrous -35 he put up last year -- and that the Capitals are going to make the playoffs.  That’s all they need to know.

To them I say it’s time to look beyond goals. Ovechkin’s offensive numbers actually aren’t particularly impressive. In 5-on-5 play he’s 90^th in the league in points per 60 minutes. Plus, Ovechkin has only 26 assists, a microscopic 14 of which are 5-on-5. Since 1970, the fewest assists by a forward to win MVP in a full season was 45 – and that was by Brett Hull in 1990-91, when he potted 86 goals. Even adjusting for the overall decrease in scoring in the league since 1991, Ovechkin’s season isn’t in the same universe,

The love from some in the analytics community is more unexpected. Their argument focuses primarily on Ovechkin’s solid shot attempt differential of 54.5%. They then apply some magic analytics pixie dust by disregarding Ovechkin’s actual goal differential to calculate what the Capitals’ “expected” goals for and against when Ovechkin is on the ice would be if Ovechkin’s linemates and goalies had a “league-average shooting and goaltending,” rather than the considerably below-average shooting and save percentages they actually have. One commentator suggested that the Capitals’ shooting percentage when Ovechkin’s on the ice should be assumed to increase by 8 points from 9.12% to 9.20%, and the Capitals’ save percentage should be assumed to increase by 14 points, from 90.9% to 92.3%.

In other words, these analytics experts are assuming that Ovechkin’s actual productivity is artificially low because of bad luck and natural variance in volatile stats like shooting and save percentage.

They seem oblivious to the possibility that his linemates’ and goalies’ struggles are actually caused by Ovechkin’s play.

Aside from being a premier goal scorer on the power play there are two things about Ovechkin’s game that stand out. First, he has never seen a shot he didn’t like. Over the past four seasons Ovechkin attempted 2,541 shots, a mind-boggling 635 more than anyone else in the league. Second, although his defense has definitely been better under new coach Barry Trotz, Ovechkin is still prone to grotesque defensive lapses.

Assuming that Ovechkin “should” be getting league average shooting and save percentages from his teammates conveniently assumes these problems away.

When a player is as poor defensively as Ovechkin is, it isn’t realistic to assume his play will have no impact on his goalies’ save percentages.

When a team’s offense runs through a player with a unique propensity to shoot indiscriminately, miss the net a ton, and rarely pass, then it isn’t realistic to assume that his linemates’ shooting percentages are going to be the same as they would be if they played with a more conventional player.

So I’m going to suggest something truly radical to the analytics community: to gauge Ovechkin’s value we should focus on actual, real-life goals.

When we look at goal metrics, it turns out that when Ovechkin is on the ice in 5-on-5 play, the Capitals’ goal differential this season is 0.8% lower than when he isn’t on the ice (this is called Goals For % “Rel.”, as in “relative” to the rest of the team). In other words, 5-on-5 the Capitals do better when Ovechkin isn’t on the ice than when he is.

For some elite players it would be fair to dismiss such a result as an aberration. Over the relatively small sample of 70 games, it would be reasonable to conclude that the player was a victim of variance; that he’s just on the wrong side of puck luck.

For example, so far this season Patrice Bergeron, Anze Kopitar, and John Tavares all fall in the same negative territory as Ovechkin.

But as shown in the table, the difference is that these players have a long track record establishing their positive goal differentials, whereas Ovechkin has just the opposite. This year’s negative numbers aren’t an aberration for Ovechkin, they’re the norm. They’re most likely an accurate reflection of his true contribution to his team.

Contrary to what the analysts are saying, Ovechkin isn’t even break-even for the Caps, let alone the league’s MVP.

GOALS FOR % 2010-2015 (5v5)

*numbers courtesy of www.war-on-ice.com

In his last three seasons as coach of the Washington Capitals (2009-2011) he won one President’s Trophy (first place overall in the regular season), and came second and fourth overall in the two others. But despite the tremendous regular season success, the Caps couldn’t manage to get past the second round of the playoffs.

Signs point to this year’s Ducks suffering the same unkind fate. Despite their dominance in the standings, advanced stats do not bode well for the Ducks’ playoff success.

As we’ve pointed out in previous columns, Score Adjusted Corsi (“SAC”) is a pretty solid predictor of long-term team success. Sometimes a team can outrun its SAC for a while—like the Avalanche did for all of last season and like the Flames have managed to do so far this year. But unless your goalie is Carey Price or you manage to have guys named Crosby and Malkin as your 1-2 down the middle, the inevitable gravitational pull of shot differential takes hold, and teams usually perform about where you’d guess based on their SAC.

That’s bad news for the Ducks, which rank 17^th overall in SAC.

To make matters worse, if the Ducks hang on to the first seed in the West their likely first round opponents would be the Wild, Jets, or Kings, who rank 11^th, 7^th, and 2^nd overall in SAC respectively. Despite finishing first overall, an argument could be made that the Ducks would actually be the underdog against any one of those teams, based on their SACs.

At this point it’s reasonable to ask how the Ducks have managed to lead the league in points despite their below-average SAC, and why I think they can’t keep it up going into the playoffs.

There are a number of factors, but the main one is the Ducks’ record in one-goal games:  26 wins, 1 loss, 7 OT losses, for a win percentage of 76.5%, which is almost unheard of. Not only is it good for first overall, there are only 5 teams in the league with one-goal game win percentages over 60%.

Now your traditional analysts will tell you that this is great news for the Ducks. It shows they have “heart” and “grit” and “know how to win.” They’re. Just. Wrong.

One goal games are a coinflip. Winning one goal games is not a skill. It is not repeatable in the long run.

Don’t believe me? Three of the last four Stanley Cup champs had losing records in one-goal games during the season in which they won the Cup. Last year’s Kings won just 48.8% of their one-goal games, and the 2012 Kings and 2011 Bruins won only 37.0% and 47.1% respectively.

Heck, even the lowly the Leafs had the seventh best one-goal game win percentage in the league last year (54.3%), and the eighth best the year before (57.9%), and let’s just say that they’re not exactly on the top of anybody’s list of gritty, battle-tested teams that “just know how to win.”

Now none of this is meant to suggest that there’s anything wrong with winning one-goal games. A team will happily take the two points every time.  But contrary to what most analysts say, it’s not a signal that a team knows how to win. It’s a signal that the team is getting lucky and winning more games than it will after the luck evens out.

And in the long run luck always evens out.

For the Ducks, it means that their point total probably doesn’t reflect the team’s true performance. It means their one-goal game record has masked their weak SAC. It means that the Ducks aren’t nearly as good as a lot of people think they are.

In fact, in games decided by more than one goal, which appears to be a much better measure of team’s ability, the Ducks have a losing record (16 wins, 19 losses).

None of this is true of the Ducks’ potential first-round opponents. The Wild and Jets are right around .500 in one-goal games, and both have solid winning percentages in games decided by more than one goal. And the Kings in this respect are the anti-Ducks. Their SAC is second overall, but they’re 28^th in one-goal game win percentage (35.5%), with a 60.6% win percentage in games decided by more than a single goal. The Ducks are overperforming; the Kings are much better than their record suggests.

None of these opponents is particularly appealing if you’re Bruce Boudreau.

Bruce Boudreau’s team could very well end up winning the President’s Trophy. But as weird as it seems, you should probably feel bad for the guy if that happens, because with the Ducks’ grossly inflated point total comes grossly inflated expectations for post-season success.

Which is bad news for Boudreau, for whom it could well be different team, same result, as there’s a solid chance that the league-leading Ducks will have a very short post-season.

In early December, when Calgary was riding a white hot 17-7-1 start, we wrote an article explaining that its success was nothing but smoke and mirrors. We pointed out that puck luck was duping just about everybody from Toronto to Timbuktu into believing that the Flames were an up-and-coming contender, when in reality they’re Michael Spinks and most of the rest of the NHL is Mike Tyson. On cue, reality set in and Calgary promptly lost its next 8 games.

So what exactly is puck luck from a hockey analytics perspective?

In a nutshell, there are certain crucial aspects of performance that tend to vary significantly in the short or even medium term (say 10 to 40 games), but which virtually always return to long term averages. “Puck luck” is just shorthand for whether a player or team is running unsustainably hot or cold as compared to those long term averages.

The most important of these performance metrics are shooting percentage and save percentage. It’s not news to anybody that players often go on hot streaks where every time they touch the puck it seems to bulge the twine, or cold streaks where they couldn’t hit the ocean from the side of a dock. But what people seem not to always appreciate is that over time these streaks end and players almost always revert to their historical average shooting percentages.

There’s even a stat specifically designed to provide a very rough measure of puck luck: “PDO.” PDO is nothing more than a single number that adds a team’s shooting percentage and save percentage. By definition the league average PDO is exactly 100 (because save percentage is the inverse of shooting percentage). Teams with particularly skilled shooters or puckstoppers (or great team defense) can maintain a PDO of 101 or even higher, but as a rule of thumb any team with a PDO of above 101.5 (or below 98.5) is probably running unsustainably hot (or cold) and can expect the inescapable gravity of regression to historical averages to take hold and pull the team back down (or up) to its expected percentages.

The Buffalo Sabres have been kind enough to provide us with a textbook example of the power of puck luck. Over their first 18 games the Sabres won 3 and lost 15. They scored an almost impossibly low 9 goals in their first 10 games, including being shut out 4 times.

But in their next 13 games they won ten times and scored 44. Had the Sabres turned the corner? Had they instituted new systems that plugged the gaping holes in their defense while simultaneously jacking up their offense to the point where it was the most potent in the league? Had they secretly lured Wayne Gretzky out of retirement and covertly named him Zemgus Girgensons just to throw everyone off the scent?

Let me put it this way: all of the above are equally plausible explanations.

As the table illustrates, the Sabres’ 13 game hot streak was all about puck luck.

BUFFALO SABRES 2014-15 (5v5)

Other than a slight improvement in Corsi Against, the Sabres’ shot metrics during their 10-3 streak were virtually identical to their 3-13-2 start. But their save percentage jumped from 17^th overall to 5^th, and their shooting percentage almost doubled, rocketing the Sabres from 28^th to 1^st (stats courtesy of war-on-ice.com). Several of the Sabres’ top scorers were shooting miles above their historical percentages. Marcus Foligno was shooting 27.3% (compared to his career average of 12.6%), Girgensons shot 26.7% (9.9% career average), Matt Moulson 19.2% (13.4% career average), and Brian Flynn 15.0% (8.7% career average).

Ironically, Buffalo’s puck “luck” might have been decidedly unlucky, since it’s only practical effect was to jeopardize Buffalo’s odds in the Connor McDavid sweepstakes.

If you’re a Buffalo fan you shouldn’t worry too much though, the hot streak was never sustainable. It wasn’t “real” in the sense that it was reflective of some actual improvement in the Sabres’ play, it was just puck luck at work. As demonstrated by the Sabres’ much more representative one win in their last nine games, your team is still by far the worst in the league and I promise it will be in the thick of the hunt for that first overall pick at season’s end.

NHL teams have finally bought into hockey analytics, so why is it that most of the media remains oblivious? As just one shining example, Sportsnet recently ran an article headlined “Flames and Oilers Heading In Opposite Directions.” That headline was pretty much the only thing the article got right, and even that was accidental.

Calgary and Edmonton are indeed two teams heading in opposite directions, but it’s not the directions the author and most other people think.

The writer predictably contrasted Calgary’s surprising 12-6-2 start with Edmonton’s disappointing 6-11-2. The article marveled at Calgary’s “character” and “courage,” ever-so-delicately declaring that “Calgary has more guts than a killing floor.”

The writer proclaimed that 20 games into its “rebuild” Calgary has accomplished more than Edmonton has in 5 years.

Tune in to the talk shows and it’s tough to find anyone who isn’t on the on the Flames’ bandwagon. On the day the article ran, Sportsnet’s power rankings had Calgary 8^th and TSN’s had them 10^th. Both had Edmonton 28^th.

So my question is this: Are the big boys of hockey media so set in their reliance on their almost mystical explanations for things they can’t explain (Confidence! Momentum! . . . ) that they’re ignoring the most basic numbers that matter, or is their fundamental misunderstanding of the dynamics of hockey so ingrained that they just can’t see past it?

Obviously wins and losses are important. But if journalists are going to definitively proclaim the resounding success of one team’s “rebuild” and dismal failure of another’s, they should at least know enough to realize that just 20 games into a season, wins and losses are a seriously unreliable measuring stick.

In this respect hockey’s a lot like poker. Both are games of skill, and in the long run the best player/team will win. But whether you want to call it “luck” or “randomness” or “variance,” as anyone who’s ever been beaten by a 2-outer on the river can tell you, in the short run factors other than skill often tip the scales, especially in a league with as much parity as the NHL (as we described in our November 6 article).

Fortunately, analytics gives us much better metrics for gauging performance. Twenty games isn’t a huge sample, but there’s enough to get a sense of what’s going on.

And what’s going on is that, despite its uncanny ability to give up prime scoring chances with disturbing regularity, Edmonton is a better team than Calgary. Don’t get me wrong, nobody’s going to confuse the Oilers with a good team, and they probably make more glaring errors than any NHL team that isn’t located on the shores of Lake Ontario. But if you gave them average goaltending they’d be almost respectable.

As of the date of the Sportsnet article the Oilers’ Corsi For % of 50.4 was almost 7 points higher than Calgary’s (43.7), which ranked third worst overall. Calgary had a shot attempt differential of -205; Edmonton’s was +59.

While shot metrics alone never tell the whole story, there’s no universe where a team with a CF% of 43.7% through 20 games has been better than a team with a CF% of 50.4%.

Flames and Oilers Team Stats (2014-15, 5v5, as of 11.20.14)

Calgary’s much better record is mostly explained by its PDO (a common measure of “puck luck” consisting of save percentage plus shooting percentage) of 103.3, which was third highest in the league. Edmonton was 29^th, at 97.4. As we explained in our October 30 column, high PDOs are sustainable – if you have off-the-charts talent like the 1983 Edmonton Oilers. But the Flames aren’t even the 2014 Oilers, let alone the 1983 edition. Still, through 20 games Calgary somehow managed to put up the very best 5v5 shooting percentage in the league, at 10.6%.

Hands up everyone who thinks that’s gonna continue.

Finally, speaking specifically to each team’s rebuild, Edmonton’s CF% is 2.7% higher than last season’s, whereas Calgary’s is 2.6% lower. Somebody needs to explain to me where all those “guts” and “courage” and “character” are reflected in those numbers.

Much like the Leafs’ early success last season, Calgary’s start this year is smoke and mirrors.

Perception is a funny thing. Many journalists (and coaches, and fans) have a hard time separating results from actual performance, probably because they don’t realize just how big a role randomness/luck plays over a 20 (or 40 or even 80) game span. Hockey analytics can go a long way toward explaining what’s really going on by providing substance and insight that aren’t limited to the tired and trite hockey clichés that many people in the hockey establishment still treat like inviolable truths.

 There isn’t much disagreement as to the first two, but most hockey fans wouldn’t put Bergeron in the same class.

They’d be wrong.

In almost every aspect of the game – both offensive and defensive – last season Bergeron not only held his own against “Captain Canada” (Crosby) and “Captain Serious” (Toews), he actually beat them or even blew them completely out of the water.

Everybody realizes Bergeron is a premier defensive player, but before we talk defense let’s consider offense.  Last season in 5-on-5 play Bergeron scored as many goals as Toews (19) and just one fewer than Crosby. Because Bergeron gets slightly less ice time than those two, his goals per 60 minutes is actually higher than both of theirs.

In terms of total points, as might be expected Bergeron fares slightly worse, though his 2.27 points per 60 minutes is just a shade behind Toews’ 2.35 P/60 and within striking distance of Crosby’s 2.54 P/60.

That’s offense. Let’s look at some metrics that incorporate defensive performance, which has always been Bergeron’s strong suit. Because Corsi For % (CF%) reflects the percentage of shots taken by the player’s team relative to his opponent, it incorporates defensive play.

There’s nobody better than Mr. Bergeron in CF%. At 61.2% last year (in 5-on-5 play) he was #1 in the league. Toews’ excellent 59.1% was good for 8th overall, while Crosby only registered 53.0% (30^th).

To be fair, CF% isn’t exactly an apples-to-apples comparison because the Bruins and Blackhawks were much better possession teams than the Penguins. So Crosby’s CF% would have been brought down by his teammates’ performance.

To account for this factor many in the hockey analytics community prefer a stat called “CF Rel%” As we explained last week, CF Rel% measures each player’s CF% “Rel”ative to that of his team. Basically it reflects how much better (or worse) the player’s CF% was relative to his team’s average, thus providing a Corsi-based stat that more accurately reflects just the player’s own performance.

It turns out Bergeron’s CF% Rel stacks up even better than his straight CF%. At +9.69 (3^rd overall), Bergeron was almost ten points higher than the average player on a Bruins team that was already one of the best in the league.  Crosby’s CF% Rel was a very solid +6.93 (19^th), with Toews at +5.33 (38^th).

Relatedly, of all forward lines playing at least 300 minutes together last season, Bergeron’s line (with Marchand, and Reilly Smith) had a 5-on-5 CF% of 62.4, which was #1 overall. Toews, with his considerably better linemates Marion Hossa and Patrick Sharp, were second overall, but still way behind, at 59.6%. And Crosby with Chris Kunitz and Pascal Dupuis were tenth at 56.5%

And just for emphasis, let’s not forget to note that Bergeron started only 46% of his shifts in the offensive zone last year (compared to 51.4% for Crosby and 63.9%) for Toews. And, speaking of faceoffs, Bergeron is one of the best faceoff men in the league, winning 58.6% last year, good enough for third overall.

BERGERON vs. CROSBY vs. TOEWS

2013-14 (5-on-5)

For the coup de grace let’s look at one of my favorite stats:  WOWYs. “WOWY” is an acronym for With Or Without You. It measures the performance of a player by looking at whether other players produce more when playing with him or without him. Captains Canada and Serious have excellent WOWYs. But Bergeron’s is jaw-dropping.

(Click on images to enlarge).

The graphs below compare the CF% WOWYs for every player that played at least 250 minutes last season with either Bergeron or Ryan Getzlaf.

Although Getzlaf also has to be in the running for best forward in the world after finishing second in both scoring and MVP voting last season, he doesn’t hold a candle to Bergeron.

Using a simple average, Bergeron’s CF% WOWY was +10.1. In other words, on average his Bruins teammates had a CF% more than ten points higher when playing with Bergeron versus playing without him.

Getzlaf’s average was +0.8.

And just to drive the point home, take a look at what Bergeron did for Zdeno Chara – arguably the best defenseman since Nik Lidstrom. It’s not that hard to take a bad player and make him look average, but Bergeron took one of the best defensemen of our generation and added a massive 8.7 points to his CF%. Unreal.

Bergeron isn’t even the captain of the Bruins (Chara is). But I think he deserves a captain nickname to go with Crosby’s Captain Canada and Toews’ Captain Serious. Given what the numbers tell us, for Begeron “Captain Everything” has a nice ring to it.

The Department of Hockey Analytics employs advanced statistical methods and innovative approaches to better understand the game of hockey. Its three founders are Ian Cooper, a lawyer, former player agent and Wharton Business School graduate; Dr. Phil Curry, a professor of economics at the University of Waterloo; and IJay Palansky, a litigator at the law firm of Armstrong Teasdale, former high-stakes professional poker player, and Harvard Law School graduate. Please visit us on line at www.depthockeyanalytics.com

Imagine the damage they could do if they had a bona fide first line centre to play with.

Chris Kunitz has Sidney Crosby. Corey Perry has Ryan Getzlaf. Patrick Sharp has Jonathan Toews. Alex Ovechkin has Nicklas Backstrom. Kessel and van Riemsdyk have . . . Tyler Bozak.

Now unless you’re one of the legion of masochists known as Leafs fans, you’re probably asking “who the heck is Tyler Bozak?”

And if you did, nobody could blame you. To say that Bozak isn’t exactly your prototypical first line centre in the mold of the superstars listed above is like saying that Tim Tebow isn’t exactly Peyton Manning.

Bozak did score 2.32 points per 60 minutes (5-on-5) last year, which was the highest among all Leafs centres. But that’s hardly enough to silence Toronto’s notoriously unforgiving fan base, who insist that Bozak’s really not much more than a passable third-liner who’s managed to hit the jackpot by getting slotted between two of the league’s premier wingmen.

The alternative to Bozak is second line centre Nazem Kadri. Kadri has shown flashes of brilliance in his four-year NHL career, but he’s also spent much of the last few seasons in coach Randy Carlyle’s doghouse because of his defensive lapses and perceived inconsistency.  Kadri scored 1.71 points per 60 minutes (5-on-5) last season.

So which of the two should be the Leafs’ #1 man in the middle?

Under old school thinking, Bozak’s considerably better P/60 would pretty much be the start and end of the discussion. But there are two problems with that approach.

First, because goals happen so infrequently, points may not be an accurate measure of actual performance unless you’re using more than a full season’s data. One way to account for this problem is to measure performance using Corsi For % (CF%) – which is the percentage of all shot attempts by both teams that are taken by a player’s team when he’s on the ice. Because shot attempts happen so much more often than goals, it takes far fewer games for CF% to provide a reasonable measure of performance, and CF% has been shown to be a good predictor of goals over the long run.

Second, regardless of whether P/60 or CF% is used, if we just compare the individual numbers of the two centres the comparison won’t be apples-to-apples because Bozak gets to play with much better linemates. A straight-up comparison of each player’s P/60 or CF% would be measuring (1) whether Bozak playing with Kessel and JVR is better than Kadri playing with second-liners. But what we really want to know is (2) whether Bozak playing with Kessel and JVR is better than Kadri playing with Kessel and JVR.

It turns out that in the 271 minutes Kadri played with Kessel and JVR last season, that line’s CF% was 49.3. In the 831 minutes Bozak played with those wingers, their CF% was only 46.2 (stats courtesy of Progressivehockey.com).

That’s a big difference, and it strongly suggests Kadri is the right man for the job.

Now 271 isn’t a whole lot of minutes, so to double-check my results I expanded the comparison to include every Leaf that Bozak or Kadri played with for 200+ minutes last year.

As the graph shows, with only a single exception (defenseman Cody Franson) every single Leaf did considerably better when playing with Kadri than with Bozak. In some cases (wingers Joffrey Lupul, David Clakson, Mason Raymond) the differences were astronomical.

This leaves little doubt that Kadri is the better player.

So the conclusion has to be that Kadri should be the first line centre, right?

Ordinarily the answer would be an unequivocal “yes.” But in this particular instance there’s a weird dynamic working in the background. The graph illustrates that definite second-liner Lupul and possible second-liner Clarkson do so much worse with Bozak than Kadri that moving Bozak onto the second line would decimate the second line’s productivity.

This confirms the Bozak-haters’ belief that his success is due much more to his elite wingers than Bozak himself.

But at the same time it cuts the legs out from the anti-Bozak camp’s argument that he should be deposed from the top line. Even though Kadri is clearly the better player, and even though Kessel and van Riemsdyk do considerably better with Kadri, the numbers show that a second line with Bozak in the middle would be so feeble that in the aggregate the top two lines’ production would be much lower.

Ultimately Bozak is lucky he isn’t a better player.  If he were, he’d get demoted for sure.

The “summer of hockey analytics” continued on Tuesday, when the Leafs reportedly hired three shiny new additions to their analytics team. In light of the hirings the Star asked the Department of Hockey Analytics what we would tell Brendan Shanahan if he called us up looking for a little analytics advice.

1. All the stats in the world won’t do you much good if you don’t understand how to use them.

I’ve been told by more than one NHL GM that the team that does the most analytics in the NHL is . . . the Buffalo Sabres. “Doing” analytics clearly isn’t the answer; you’ve got to do it right.

Analytics is much more than gathering data and obsessively focusing on a few new stats. Analytics is a method to better understand all of the complicated, interrelated aspects of the game of hockey.

What most teams fail to understand is that statistics are the input, not the output. Each stat is a piece of a complicated puzzle. Each has shortcomings, limitations, and correct and incorrect applications. Analytics is the process of piecing the puzzle together, usually requiring the creative application of sophisticated statistical tools. It’s easy to get lost in the numbers, or even to just misinterpret what the numbers are telling you.

To use just one example that we wrote about earlier this year, “super-sniper” Alex Ovechkin actually shoots an awful lot of blanks. When we looked at him mid-season his shooting percentage over the prior three years was a gaudy 13.4%. But Ovi misses the net a ton, so his “true” shooting percentage (goals divided by total shot attempts) was only 6.74% -- third worst among the 27 elite scorers we looked at. Interesting, sure, but what do you do with it? The stats start the process; but then you have to ask the right questions and have the tools to answer them. Why does Ovi miss the net so much? What are the effects of his misses? How often do they result in turnovers? Does his “shoot now ask questions later” approach help explain his linemates’ dismally low shooting percentages? Is there some tactical change that could mitigate the effects of his misses while still making use of his cannon of a shot?

Answering these questions and figuring out how to use (or stop) Ovechkin takes a lot more than a huge database full of numbers. Don’t be the Buffalo Sabres.

2. Possession isn’t everything. The vast majority of people doing analytics focus predominantly on possession and shot metrics like Corsi and Fenwick, or derivations of them, like zone entry stats. Don’t be fooled: those are just the tip of the iceberg. The reality is that many in the analytics community focus on those stats because they’re easy to understand and use. They prefer a tidy world where a shot attempt is a shot attempt and they don’t need to worry about pesky issues like the quality of the shots that are taken. You need to make sure your analytics guys can handle these types of questions.

3. Fire Carlyle. Possession isn’t everything, but it’s pretty darn important, and Randy Carlyle’s system and puck possession go together like nuts and gum. As soon as Carlyle showed up the Leafs went from bad to worse and then from worse to almost impossibly awful. Things were no different in his last years in Anaheim. The table shows the Corsi For percentage (the percentage of shot attempts taken by a team relative to shot attempts by their opponents) of the Ducks and Leafs, with Carlyle’s years in bold. As soon as Carlyle left the Ducks, their numbers jumped. Not at all coincidentally, as soon as he landed in Toronto, the Leafs’ numbers fell off a cliff.

Randy Carlyle’s Teams

Corsi For % (5v5)

4. Educate the organization. The Leafs organization is officially schizophrenic. New school analytics meets old school coach and GM. Probably not a recipe for success, but Shanahan can smooth the way by imposing a structure that will at least help the new crew explain to the rest of organization what analytics is and how it can help them do their jobs. Despite what Carlyle and Nonis may think, analytics isn’t voodoo, it’s just different information. More information is always better (as long as it’s useful and it’s right). The more Shanahan can get everyone on the same page, the more coherent a plan the Leafs will be able to execute, from team composition, to roster, to lineups, to on-ice tactics.

5. Get ready for big data. Within the next two seasons the NHL will likely adopt SportVu or some other system that continually tracks the movements and actions of every player on the ice. That will provide raw data unlike anything ever seen in hockey in terms of both quantity and content. Overnight there will be accurate information about, well, everything: zone stats, scoring chances/shot quality, puck retrieval, player speed, passing stats, etc., etc.  The sky’s the limit.  Much of what’s come before will be obsolete. The Leafs will need to establish a system to organize, search, filter and analyze the tidal wave of data. As stats guru Nate Silver emphasizes, the ability to analyze the stats and use them in creative ways to gain a competitive advantage will become paramount once big data hits the NHL.

Last week we pointed out that a team looking for a steal in the free agent market should snap up Radim Vrbata if they could get him for anything like his 2013-14 salary of $3,000,000.

Benoit Pouliot might be an even better buy.

At first glance, Pouliot’s numbers are solid but unexceptional. Over the past three years he’s scored 39 goals and 49 assists playing for the Rangers, Bruins, and Lightning. But those humdrum stats belie some pretty impressive productivity.

Because he played for some deep teams, over those three years Pouliot averaged less than 13 minutes per game. So when we normalize his productivity to account for differences in playing time, we see that he’s actually more productive in 5-on-5 play than almost any other player in this year’s free agent crop, including marquee free agent and 6.6 million dollar man Paul Stastny.

*Jagr is included because he was part of this year’s free agent class, though he re-signed with the Devils soon after the end of the season.

Pouliot stacks up well in terms of production per 60 minutes, but when we also account for the fact that his cap hit last year was only $1.3 million, Pouliot blows the competition out of the water. If we divide salary by the number of goals, assists, and points, we see that the Rangers got about four times as much bang for their buck out of Pouliot last year as the Avalanche got out of Stastny. Pouliot’s goals/assists/points per dollar were also between double and triple those of Jagr and Mr. Rent-A-Disappointment, Thomas Vanek.

When we add in the fact that Pouliot is also a solid defensive player (with an average goals for percentage above 60 over past three years) who has developed a physical aspect to his game (a hit differential of +58 in 2013-14), he looks even better relative to some of the uni-dimensional free agents on the market.

All of that should be enough for a team to jump at the chance to sign Pouliot. But to me there’s one other set of statistics that might be even more impressive. A useful tool for assessing a player’s individual contribution to his team’s success is to measure whether other players on his team perform better when playing with him or without him. These statistics are called “WOWYs” (“With Or Without You”). These stats measure whether the player’s performance improves or detracts from the productivity of his teammates, which may provide an indirect measure of aspects of his overall contribution that might not be picked up in point totals or even possession metrics like Corsi or Fenwick.

The final graph shows Pouliot’s WOWYs for Corsi For %, which measures the percentage of shots taken by his team as opposed to the opposition when he’s on the ice.

Remarkably, every single player to play substantial minutes with Pouliot last year had a better Corsi For % when playing with him than they did when playing without him. Some players (e.g., McDonagh or Richards) may have faced tougher opponents when playing with linemates other than Pouliot, and it’s also possible Pouliot just had a lucky year. But guess what? Almost every player Pouliot played with in Boston and Tampa during the two previous seasons also had better Corsi For percentages when playing with Poutliot than when playing ithout him. That tells us this guy is doing something that makes his teammates better.

Pouliot came into the NHL as a 4^th overall pick in a year in which the 4 other top 5 picks were Sidney Crosby, Bobby Ryan, Jack Johnson, and Carey Price, and other stars such as Anze Kopitar, Paul Stastny, James Neal, and Tuuka Rask were all picked after him.

That’s tough company to keep up with, but at a cap hit of $1,300,000 last season, this is a guy who offers reliable skill at a fraction of the price.

Combine all of that with the possibility that Pouliot may be overlooked by some teams because he had a slightly below average year last year, and there’s a real chance that he’ll be a bargain buy for some smart team.

But if they did, who should win this year?

Let’s start with the assumption that the Conn Smythe winner will come from the team that wins the Cup.

If the Rangers win . . . just kidding, the Rangers aren’t going to win.

If (when) the Kings win, there’s an interesting group of candidates. Most commentators seem to think that Anze Kopitar and Drew Doughty have the inside line, with Marian Gaborik, as the leading goal scorer in the playoffs, having an outside shot.

But as you might have noticed I tend not to put too much credence in what “most commentators” say. So let’s count down my top 5.

5. Marian Gaborik: Gaborik is an important piece of the puzzle, but he’s not the main guy.  He’s tops in goals scored, but otherwise not in the same class as the others on this list. His 5-on-5 Fenwick For percentage (the percent of all unblocked shots attempted when he’s on the ice that were taken by his team rather than the opponent) was just 50.2% — 15^th on the Kings and 60^th overall. And that’s playing the majority of his minutes with top-notch teammates Kopitar (52.9% of his minutes) and Doughty (33.5% of his minutes). Great playoffs, but not Conn Smythe material.

4. Drew Doughty: Doughty’s one of the best D-men in the game and he’s critical to L.A.’s success, but playoff MVP?  Not this year. True, he plays huge minutes: 28.2 per game. And true, he has 17 points.  But that number is less impressive when you consider how much ice time he gets: His points per 60 minutes in 5-on-5 play is an unexceptional 0.97 —  10^th best among defensemen in these playoffs, and just over half of Brent Seabrook’s 1.92. Combine that with the fact that a very high 37.3% of his shifts have started in the offensive zone, and the fact that he tried to hand the Blackhawks a couple games in the Conference Finals with giveaways that make Oprah Winfrey – of the “everybody gets a new car” specials – look like Ebenezer Scrooge, and I’m going to say that giving the award to Doughty would be a major misfire.  Maybe next year.

Which brings us to the three guys who have been a cut above. An argument could definitely be made for any one of the three. In fact, if one of these players has a big game from here on out he’s probably going to be the one to take home the hardware. But if the award were handed out today . . .

3. Jeff Carter:  Carter is tied for second in the playoffs in scoring with 10 goals and 14 assists for 24 points, and is just a fraction behind Kopitar with 2.77 points per 60 minutes of 5-on-5 play. He’s also second among Kings forwards in 5-on-5 Fenwick For %, at 55.9. His numbers might be slightly inflated because 38.4% of his shifts started in the offensive zone, but for the possession-monster L.A. Kings that number isn’t all that much higher than most of the others on this list. There’s nothing negative to say about his play, he just wasn’t quite as good as my top 2.

2. Anze Kopitar:  This guy’s the real deal. He leads the playoffs in scoring with 26 points, and is 6^th overall with 2.83 points per 60 minutes of 5-on-5 play. An elite defender who gets more than his share of defensive zone starts and who’s won 53.1% of his draws to boot. The one knock against him is that he’s only scored five goals in 25 games, which isn’t a mortal sin, but also isn’t enough to put him over the top.

And the Conn Smythe winner should be . . .

1. Justin Williams. You’re probably surprised. Frankly, so am I. And if I were a GM and I could have my pick of players on the Kings Williams likely wouldn’t crack the top 10. But for the past 25 games Williams hasn’t just been “Mr. Game 7,” he’s been the Kings’ most productive player. He’s tied with Carter for second in playoff scoring with 24 points, but because he plays considerably fewer minutes — especially on the power play — his 5-on-5 points per 60 minutes is much higher. In fact, Williams’ 3.33 points per 60 is by far the highest in the league. Only 2 other players in these playoffs were even above 3.00:  Evgeni Malkin and Jussi Jokinen. Williams’ 53.1% 5-on-5 Fenwick For % is better than Gaborik (50.2%), Kopitar (51.6%), and even Doughty (52.5%), and that’s despite (a) playing mostly with some of the Kings’ lesser-lights like Jarret Stoll, Dwight King, and Slava Voynov; and (b) starting far fewer of his shifts in the offensive zone (31.6%) than any of the others on this list.

Williams’ numbers are better than the other candidates and he’s delivered in critical situations and while playing with weaker teammates. That’s why he’d get my vote for the Conn Smythe.

First, even though the data set included all Finals and Semi-finals since 1980, it's still a relatively small sample.  That said, the GFDiff relationship was significant at the 1% level and the GADiff was significant at the 5%, so the numbers appear solid.

Second, at some point I'll add in data for all playoff series and/or for all teams and we'll see how that affects things.  And/or run separate regressions for the early playoff rounds or the regular season.  If I had to guess, I'd guess that if anything the GFDiff coefficient would increase if we added regular season and early playoff rounds because by-and-large teams reaching the final 4 are going to be stronger defensively and strong offenses are more likely to thrive against weaker defensive teams.  Of course, the reverse is going to be true as well.  However, because teams with strong defenses but not-so-strong offenses almost by definition play a lower-event style, they're going to be more susceptible to short-term randomness -- think the NJ Devils at the end of 2012-13 when they just couldn't buy a goal even though they played very good defense.  (Another way to think about this is that more offense generally means more opportunities, thereby reducing game-to-game variance).

Third, it would be interesting to run the numbers for just the post-salary cap era.  The problem with doing that is sample size.  But in light of the fact that teams generally pay much more for offense than defense, it occurs to me that putting together the type of team that can offensively dominate is likely more difficult, and almost certainly more difficult to sustain over multiple years, than it was previously.  Defense is going to be easier to repeat (e.g., it's not subject to the high variance that shooting percentage is), and I expect it's also easier to maintain under the salary cap.  From that perspective, then, I wouldn't be surprised if a really good defensive team might not have a slightly lower probability of winning the Cup in any single year, but would be more likely to maintain its performance, and so over a period of several years in the aggregate it would have a better overall probability of winning the Cup than a team that has huge offensive numbers for a year or two and then either (a) regresses to the mean and/or just has an off year, or (b) is forced to re-tool for cap reasons.

With that, here's the article:

“Defense wins championships.”

This might well be the most widely used truism in sports. Hockey, football, basketball, soccer: “Defense wins championships.” It’s so ingrained that it’s probably the mantra of checkers players too.

Implicit in this seemingly universal belief is that in the last rounds of the playoffs, when competition stiffens, offense may fizzle or be neutralized, but stout defense will hold the line and carry the day.

Everyone seems to believe. But is it true?

Obviously there’s something to it. Teams that treat their nets like storage lockers for their opponents’ hockey pucks aren’t likely to get too far in the playoffs.

Then again, as reflected in the table, it’s not as if the team with the best defense wins the Cup every year. Far from it.

Since 1980, only five teams that finished first overall in goals against in the regular season won the Cup. There were exactly the same number of Cup winners who finished first in goals for. So between offense and defense being the ticket to playoff glory, that’s a push.

But surely even if they weren’t first overall, virtually all Cup champs have been strong defensively, right? Not so.  It turns out that since 1980, out of 33 Cup winners seven were tenth or worse in regular season goals against. Sure, four of those teams happened to have players named Gretzky, Messier, Kurri, Coffey, Lemieux, Jagr, and Francis. But hey, those still count.

On the other hand, only four teams that were tenth or worse in regular season goals for won the Cup, suggesting that having a strong offense might well be more of the recipe for success than stifling defense.

That’s all fine and dandy, but it’s hardly scientific. Which means it’s time to put the fancy in fancy-stats.

A regression analysis is a powerful statistical tool for estimating the relationship between variables; for example, how team defense affects the probability of winning a playoff series. I ran a species of regression called a “probit” to estimate the impact of team defense and team offense on the chances of winning the Cup.

Team defense was measured using regular season goals against. Team offense was measured using regular season goals for. For the two teams playing in each Finals and Semi-finals since 1980, I calculated the difference between them in goals for and goals against to determine each team’s defensive and offensive strength relative to the other.

For example, in 2012, the last full NHL season, the Kings played the Devils in the Finals. The Kings scored a paltry 194 goals that year, but allowed only 179. The Devils scored 228 and allowed 209. So the Kings’ “goals for differential” was -34 (their 194 goals minus the Devils’ 228). In other words, the Devils’ offense was 34 goals better than the Kings’ that year. The Kings’ “goals against differential” was -30 (179 minus 209), meaning the Kings allowed 30 fewer goals than the Devils.

Using teams’ goal differentials, the probit regression estimated the contribution of team defense and team offense to the outcomes of each of the 99 Finals and Semi-finals series since 1980.

The results: Team offense actually had a slightly greater effect on the likelihood a team would win. Each additional goal for above the opposition increased a team’s probability of winning the series by 0.46 per cent.  For every goal allowed fewer than the opposition, a team’s probability of winning increased by 0.37 per cent.

Those numbers might not sound particularly big, but the differences in goals for and goals against between teams in the Finals are often 30 or more, and even reached as high as 125 in one instance (goals for differential between the 1983 Oilers and Islanders).

What this tells us is that when it comes down to crunch-time, it doesn’t matter much whether a team is an offensive powerhouse or constructs the stingiest of defensive shells. Score lots. Allow few. Both are almost equally good. What matters, at least as far as team offense versus team defense is concerned, is plain old goal differential.

* Note: Due to the fact that since 1980 NHL seasons have been 48 games, 80 games, 82 games, and 84 games, all goal totals have been normalized to an 82 game season to allow an “apples-to-apples” comparison.

Regression Results