# Getting Familiar With Hockey's Advanced Statistics

There's something of a renaissance going on in today's NHL. It's actually been taking place for a few years now, but is certainly coming more into focus as of late. All signs point toward the incorporation of more advanced statistics when it comes to how we gauge hockey, and how organizations operate.

While advanced statistics aren't in their infancy, we're learning more and more about different ways to incorporate these stats into hockey's puzzle. But there are plenty of things we already know, and while there are dissenters when it comes to the place of said statistics, it's abundantly clear these numbers not only have a place in hockey, but are quite important.

The following are terms you'll likely come across, what they mean, and how exactly they fit into the Bruins larger scheme.

Corsi is a statistic used for measuring puck possession. Why is that important? Well you can't score unless you have the puck, and Corsi shows what teams and players are effective in driving possession, and which are not very good.

Calculating Corsi is quite simple at both a team and individual level. Often expressed as a percentage (CF%), Corsi is a team or player's shot differential. It accounts for all shot attempts, meaning those on goal, those that are blocked, or those that go wide. Taking that sum, and juxtaposing it against how many shots against a team or player is the on the ice for is how Corsi is calculated.

A major reason there's been such an emphasis on resurrecting the Bruins fourth line this offseason was its inability to drive possession.

Taking Corsi a step further is relative Corsi, generally express as CorsiRel. What this statistic tells us is how much better (or worse) a team's overall Corsi is when a specific player is on the ice. The Bruins, as a team, are very strong in possession, which can mask a bad player's inability to hold onto the puck. The converse of this would be a good player on a bad team, who's numbers are negatively effected by his team's overall poor play.

A positive Corsi rel indicates that when said player is on the ice, the team is that many points above its average Corsi than when he is not. When a player has a negative Corsi rel, it means a team is below its own average.

Of the 435 skaters who qualified last season, Gregory Campbell, Shawn Thornton, and Daniel Paille fell in the bottom 15 in the league in CorsiRel a season ago, according to Extra Skater. (Recently taken offline, Extra Skater provided a wealth of fancy stats for the hockey community.) Their overall Corsi percentages weren't in the same stratosphere, which just goes to show how the unit struggled to be productive.

On the opposite end of the spectrum is Patrice Bergeron. While the Bruins were third as a team last season in overall CF% at 55, only behind the Blackhawks and Kings, Bergeron not only finished as the top skater with an individual CF percentage of 61.2, he finished as the top forward and second player in the league with a CorsiRel of 9.7 percent.

There are a few trends to dissect there. The Kings and Blackhawks have been fancy stats darlings the past few seasons, and have also won the last three Stanley Cups. That's no coincidence. The Bruins have also been very strong in many of these advanced metrics, which no doubt played a large role in their recent success.

From an individual perspective, it's incredibly impressive that while the Bruins were so effective in possession last year, they were even that much better when number 37 was on the ice.

The last wrinkle to Corsi is when we view it, and you'll often see it expressed in 5v5 close situations. What this refers to is when both teams are tied or a goal apart in the first two periods, or when the game is tied in the third period. The reason for using these situations is to eliminate score effects. Teams obviously change their strategies based on the numbers on the scoreboard. A team that's comfortably ahead has fewer reasons to be aggressive, and log shots. So to get the most accurate reading on a team's effectiveness in possession, we pool from when the teams are playing 5-on-5, and the game is within reach.

***

Zone starts give us a glimpse at where teams deploy players for faceoffs. What this tells us is how accountable a coach thinks a player is. If a defenseman has a very low defensive zone start percentage (expressed at DZS percentage), the coach probably doesn't trust his defensive abilities too much.

Taking that a step further, Torey Krug had a 23.8 DZS percentage last season, the lowest of any Bruins defenseman, or any regular skater besides Thornton for that matter. In his 57 games played, Chris Kelly led the Bruins with a DZS percentage of 38.6, while Zdeno Chara led all defensemen at 33.1 percent.

We'll keep it to one more advanced statistic for today.

PDO, a number often likened to "puck luck." What does it stand for? Well it's not quite an acronym, but what PDO measures — at both an individual and team level — is a combined shooting percentage and save percentage. The idea here is that, like virtually every statistic, this number will regress. When added, the mean of the two statistics is about 100. So teams below that line are likely due for an increase, while teams above that number may expect to regress back to it.

It's a number that, in a larger context, can show how sustainable a team's play is. For example, the Toronto Maple Leafs made the playoffs in the 2012-2013 season with the league's highest PDO (103 percent), while posting horrid possession numbers, as they came in dead last in CF percent. Despite struggling in puck possession, Toronto benefited from a very high shooting percentage and steady goaltending. But as those numbers regressed (see: last season), Toronto plummeted, and the team missed out on the playoffs entirely.

The Bruins are a team that generally has a pretty high PDO, which is anchored by consistent goaltending. Boston's overall save percentage has been in the top five of the league for the past four years, which indicates more of a trend than an anomaly.

On an individual level, PDO can predict similar future results that it does on a team level. If a single player has a particularly low PDO—take Nail Yakupov on Edmonton last season—he may be due for a hike in numbers. Yakupov's PDO of .952 a year ago was a factor of a low shooting percentage (only 9 percent) compared to the 20 percent he shot during his rookie campaign. Statistically speaking, he probably falls somewhere in the middle.

For the Bruins, it makes it a bit trickier when discerning info based on PDO. While many skaters came in at above 100 percent in 2013-14 season, that's again a factor of steady goaltending. One player though who could be due for a bounce back year of sorts is Loui Eriksson. With a career shooting percentage of 13.6, Eriksson's clocked in at a measly 8.7 percent last season, nearly five points below his career average. Injuries were obviously a factor in limiting Eriksson's production in his first season as a Bruin, but so was an uncharacteristically low efficiency when it came to his shots.

***

One of the most fascinating things about advanced statistics and how they're applied to hockey is this is still an evolving field. While we've come a long way in using these metrics to evaluate hockey, there's still room to grow. And that evolution could be taking place soon, as the league is reportedly looking into testing player-tracking technology, similar to what the NBA currently does, to boost its knowledge and understanding of the game. But there is sort of an exciting element here, as we're gradually gaining a better understanding of hockey, with more to come.