clock menu more-arrow no yes mobile

Filed under:

Sunday Spotlight: The Great Debate of Advanced Statistics

With all of the hype over the internet and the recent bump of certain TSN reporters shooting themselves in the foot, we decided to let our own rookie take a look at what seems to be so controversial about analyzing and utilizing stats.

If you buy something from an SB Nation link, Vox Media may earn a commission. See our ethics statement.

So, how many people actually care about these things?
So, how many people actually care about these things?
PetBugs (@petbugs13) -
[Editor's Note: This is not an educational session about advanced statistics, rather an opinion about the subject of one of our writers. If you wish to learn more about advanced stats on SB Nation, please check out some of the recommended links at the bottom of the page.]

Here we go again.

Every five or so months, some random persona approaches the mainstream media or a team's brain trust and asks the magical question, "Do you use advanced stats?" It's as if that question is the bane of the hockey world's existence, as seemingly a thousand different articles get written on the subject and a cornucopia of hockey personalities on Twitter go at it for hours.

Honestly? Enough is enough.

I can include various examples of guys like Gord Miller of TSN who paint themselves into a corner with every tweet they throw out about advanced statistics. Miller said via Twitter, "Another issue with hockey analytics: even if you think they work, do they have a place on a broadcast? Do people want an algebra lesson?"

The last sentence is what poses the biggest problem. Lesson number one about joining any sorts of media outlet, blogs included: never make any sweeping generalizations about your viewers' opinions, intelligence, or willingness to learn.

If you have read one of my articles here or game recaps, chances are you're not going to see the "algebra lesson" that Mr. Miller is talking about. More than likely, one of the posts will contain some general talk about puck possession, shot attempts, or general scoring chances.

All of these are part of advanced statistics, but there is nothing advanced about them. It's all simple addition, subtraction, recorded times on a stopwatch, and lots of excel sheets. Unless there is any calculation of areas under integrals (good luck with that on a shot attempts diagram) to determine time taken, there's no complex mathematics.

Even then, we already know that variable under the curve because it's the time we've been sitting at our desk with the stopwatch. Every regular season game contains no more than 65 minutes of recorded advanced statistics, no calculus needed.

So, why all the fuss?

It can be argued that some parents want their kids to learn more about game strategy and which routes to take on the ice because that's how they learn the game. Yes, as a catcher in baseball growing up, my father had me glued to the TV set watching Mike Piazza and Jorge Posada command the diamond. Pretty soon, my game resembled that of Jorge's right down to the catching stance, play calling styles, and my foray into the Spanish language.

Is some kid going to learn French by watching Daniel Brière and the Canadiens on a Saturday night? No, he won't. Will he learn how to tame his peach fuzz in the fashion of a George Parros mustache? He can probably sharpie one on to the dismay of his two parents sitting on the couch watching Homeland.

However, a parent wants his kid to watch breakdowns of how the game is played based on watching their favorite NHL players, not watch some guy in Movember with a handlebar mustache teaching "math."

Kids are like sponges, and hockey is a family game and pastime. Perhaps broadcasters do not want to ruin that affect in any way, and to some extent, they're correct. Nobody needs to see nine minutes of possession numbers thrown at them, but they do play an extremely important role in the overall picture. The argument that stats will improve already sub-par broadcasts is irrelevant. People come to watch hockey, not to see Jeremy Roenick or Mike Milbury argue it out. They'll make fun of Don Cherry, sure, but they're not necessarily watching him for the reasons that come across as advertised.

In Detroit, one of the top possession teams in the league as well as the most consistently competitive (not a coincidence), is known to have extreme confidence on the puck and often humiliates the other teams in that department. Other top teams, as mentioned in previous articles, include Chicago, Los Angeles, St. Louis, New Jersey, and Boston.

Notice, these teams have been consistently in the picture to win it all over the last few years, and they all happen to be at or near the top of the overall Corsi ratings, which measure shot attempts along with puck possession time. It should also be noted that only one of the aforementioned teams has not made a Finals run over the last five or so seasons, but they're pretty close.

However, there are people who need to wake up and smell the coffee, like Dave Nonis of the Toronto Maple Leafs. The Leafs made the playoffs last year thanks to a lot of luck, an extremely good top six, and an unrealistic shooting percentage. A few of my Leafs fan friends argue that it was only half of a season which didn't leave time for the usual March collapse, though I'll give them a break there. I can recall plenty of games where the Leafs were oushot by a gigantic margin and somehow won because of a hot goaltender or a fantastic goal by Nazem Kadri, James van Riemsdyk, or Phil Kessel.

The PDO bear doesn't like statistics very much. (Graphic by Josh Howard - JHowardDesign)

"We're constantly trying to find solid uses for it," Nonis told a room of reporters, including James Mirtle at a sports analytics panel at the PrimeTime Sports Management Conference in Toronto. "The last six, seven years, we've had a significant dollar amount in our budget for analytics and most of those years we didn't use it."

"The biggest thing we use is going to watch a player play," said Nonis. "I haven't seen anything that's going to stop that from being the primary source of our decisions."

Those quotes spurned several articles north of the border, including those by James Mirtle of The Globe and Mail as well as Chris Johnson of SportsNet, both of whom took in the panel in Toronto.

Nonis acknowledged that despite the fact that his team is getting torched in the shot (and Corsi) department, he's still winning games, which takes us back to Gord Miller and his Bill Parcells approach of having a team's record display what it is, and nothing else; it's not a good approach.

Personally, as the new kid on the block, I've only gotten into advanced statistics by following guys like Todd Cordell on Twitter and looking at the multitude of advanced statistics websites that have been emerging out of the blue, including ExtraSkater, whose name you may see at the bottom of plenty of recaps.

Graham Hathaway prefers the casual approach of beer, hockey games, and Microsoft Paint, in this graphic displaying FUNCLOSE%.

However, hockey's advanced statistics, unlike the advanced stats of other sports, are relatively new. Baseball's advanced statistics of batting averages, on base percentages, and ERAs have been around for nearly the entire lifespan of the sport. They're used at every single game and show up on the scoreboard next to every single player. That's not displayed at an NHL Arena for a multitude of reasons, the principal one being that every arena tracks these advanced statistics differently, a huge confounding variable, which by definition, is a lurking variable that affects the results.

For instance, a shot counter in San Jose might record a blocked shot or an attempt differently than one in Tampa Bay. It's easy for a pitch counter in baseball because he has all the time in the world, but hockey is easily the fastest mainstream spectator sport in North America, something that requires its statisticians to act instantaneously and then move on to the next shot attempt.

Another problem is sample size. Without giving a whole lesson on p-values and null hypothesis, we tend to need a large sample size to conduct any meaningful test, as the results we get often can be by chance or due to selection bias, which is a statistical bias in which there is an error in choosing the individuals or groups to take part in a scientific study. Dustin Brown or Logan Couture will always have better shooting percentages than Drew Miller or Patrick Eaves, as long as they're all at the NHL level.

The statistics have value. The Leafs and their obscene winning percentage are the exception to advanced statistics, not the rule. Everyone has that gut feeling that they're shooting at an unsustainable percentage (because they are) and are due for a correction. I cannot personally guarantee that it will happen in the next few weeks or months, but it will happen.

Sure, great possession teams are the same ones that headline guys like Pavel Datsyuk, Patrick Kane, Jonathan Toews, David Krejci, Anze Kopitar, Alexander Steen, Travis Zajac (?), Jaromir Jagr, and so on and so forth. However, it is really no coincidence that stats are the way they are. A very good possession player is going to have very good numbers. The forward who can make move after move and blow by everyone on the ice is going to have great numbers because he's holding onto the puck.

It all comes down to the team, and the concept that we'll go back to of confounding variables.

A team like the Pittsburgh Penguins, who last year had a ridiculously loaded roster, placed 19th overall in Corsi. However, they had such immense skill that it did not matter. They also had a lethal powerplay, some pretty good defense, and a fairly good penalty kill. There are players like Evgeni Malkin and Sidney Crosby, whose possession numbers are through the roof, but they often pass off to line mates who almost never hold onto the puck, like James Neal, who chooses to roof a shot instead.

Pavel Datsyuk is 20th in the league in Corsi rankings.(Photo by Dave Sandford Getty Images)

Good possession teams have good possession players across the board, but having good numbers does not guarantee success. At the end of the day, momentum swings skew data, as does player skill and overall team chemistry. The New Jersey Devils, for instance, were horrible last year after the ten game mark and managed to be the second best team in Corsi with a +412 at the end of the season, yet they missed the playoffs by a mile.

A team can be excellent in possession but continually lose because it cannot bury its chances. It is not puck luck. A team with more skill or a hot player will bury its chances time and time again. Goals are not scored randomly, unless you're the Toronto Maple Leafs.

The fact is, a hockey team wins by scoring more goals than an opponent. A baseball team can outhit another one 12-1, but if that one hit was a home run and the other team didn't have a single one of those runners cross the plate, they lose. The electoral college is what determines the national election. If a president wins the popular vote, odds are he will win the electoral college, but it doesn't always work like that, as in the year 2000.

Most teams who out shoot their opponents often will genuinely win more games, and to out shoot their opponents, as a prerequisite, they have to be in possession of the puck, hence the numbers game.

At the end of the day, the Bill Parcells approach is not always true, nor do advanced stats tell the whole story. Both of them create a great picture, sure, but it takes a little bit of intuition to truthfully build a contender. This is why scouts have jobs. If a team was built and a draft was conducted by numbers alone, things might get a little boring and a little ugly.

It's the same exact reason as to why not everybody who attends a high end University has the same GPA, same courses on his or her transcript, played Tuba or Clarinet in high school, founded a charity that feeds the homeless, started a business, or is preparing to become a Lawyer. If everybody who went to school became a lawyer, we wouldn't be a functioning society today. (Different topic for another day.)

GMs and scouting staffs take a player based on what he can bring to the table. If a GM needs a puck moving defenseman, he'll go ahead and get one. Is he going to be Shea Weber or PK Subban? Maybe not. It's more realistic to see a blue line become the some of its parts than a bunch of extremely talented defensemen.

A team is the sum of its parts, not one giant coincidence. Statistics help us determine whether or not these parts are adequate and they provide a very good medium into the grand scheme of hockey. However, they are far from perfect and they should not be used to tell the whole story.

As long as statistics are used appropriately, the world can relax. Statistics have a place in the hockey world and as some standardization across the league ensues, things will get a bit easier for everyone. For now, let's not get too carried away.

By the way, the best possession player by Corsi in the NHL last year was Justin Williams of the Los Angeles Kings.

Here are some really good reads and resources that illustrate the importance as well as misuse of advanced statistics: