Fancy Stats

Some not so fancy stats

The movie "Moneyball" (based on a book by Michael Lewis) shed a new light on how baseball teams manage and create their rosters. Smaller market teams are unable to spend with the "New York Yankees" and the "Boston Red Sox" of the world so they need to ensure that every dollar they spend is in fact worth spending.

The movie shows how the statistics like stolen bases, runs batted in, and batting average (that are typically used to gauge a player's value) are flawed and do not express a player's true value. The book/movie argues that Oakland's management took a more analytical approach to find players that would not only provide the most bang-for-their-buck but also compete successfully against the richer teams in the MLB.

The "Moneyball" story applies to the NHL as well.

Over the last six years, there has been an increasing interest in how advanced statistics apply to the game of hockey and one website has come out on top as the leading source of this information.




Behind the Net is a website that looks not only at advanced team statistics but also individual player statistics.

I could sit here and try and explain them to you, but I'll let the experts themselves do it. From Behind the Net:

First things first: traditional NHL statistics don't tell us very much about a player's true value. Rob Brown had 49 goals in 68 games in 1988-89? Playing with Mario Lemieux had something to do with it.Zdeno Chara averaged a +20 plus-minus over the last ten seasons, but was -21 in 2006-07? Might have been a reflection of his teammates. This site cuts through context to find the true skill levels of teams and players.

The core of advanced hockey analysis is shot differential statistics, the best-known of which is "Corsi," an eponymous statistic coined by blogger Vic Ferrari in 2007 and completely dismissed by Don Cherry (video) in 2010. Jim Corsi, the goaltending coach of the Buffalo Sabres, keeps count of all shots directed at net (including blocked shots and shots that miss the net.) Among the statistics that are currently collected by the NHL, this is the single-best predictor we have of future team winning percentage. Studies from past seasons or single games have also shown that It's also a proxy for things like scoring chances, puck possession and zone time. It needs to be said that these aren't new ideas: Harry Sinden did very similar things when he was a General Manager and what we do here is a re-purposing of these ideas. 

A Guided Tour through Advanced Statistics on Behindthenet.ca

Let's start with team statistics. First, Goal and Shot Rates at 5-on-5 - you can find data for all other strengths on the same page, but we focus on 5-on-5 because roughly 3/4 of NHL games are played at 5-on-5. A team that dominates shot differential at 5-on-5 tends to have very good playoff outcomes, for example, Detroit in 2007-08 or 2008-09, Chicago in 2009-10and Vancouver in 2010-11. Shot differential is obviously not the be-all and end-all of team performance; we'll get back to that later. It's also important to keep in mind that shot differential exhibits a lot of what are known as score effects: the later we get into a game, the more shots the trailing team takes and the fewer the leading team takes as it protects its lead. To account for that, we also track Fenwick (Corsi excluding blocked shots) by game score, and in what's known as "close situations" - tie games and in the first and second periods of games within two goals. The best "Fenwick Close" score that any team has achieved in the last five seasons is Detroit's 59.4% in 2007-08.

Equally important is the concept of PDO, which is simply the sum of save percentage and shooting percentage. In the long-run, PDO regresses very heavily to 1000; rare is the team that is able to stay consistently above this level, though every year, people claim that some team can out-finish their opponents while failing to out-shoot them. Examples include Colorado in 2009-10, Dallas in 2010-11 and Minnesota in 2011-12 - all three teams came crashing down to earth way before the season ended. When you want to know how a team is going to do going forward, out-shooting is way more important than out-finishing.

Individual Player Statistics

These team-level statistics have an analogous player version, like 5-on-5 Corsi (for example, Detroit in 2007-08). The league leaders in Corsi since 2007-08 are Tomas Holmstrom (+21.0 per 60 minutes), Pavel Datsyuk (+20.3) (2007-08 Game Chart) and Henrik Zetterberg (+18.2) (2007-08 Game Chart); Alex Ovechkin (+16.9) is #4. (If we adjust individual Corsi relative to a player's teammates, Daniel Sedin is #1 in the league at +16.6.) You can find links to every player's game charts here.

Unlike team-level stats, individual stats are also driven by how a coach chooses to use a player. Here, the two most-important metrics are Offensive Zone Starts and Quality of Competition. It should be obvious that players who get sent out for a lot of offensive zone faceoffs are more likely to direct a shot on goal than surrender one, and Offensive Zone Start percentage captures this. The Vancouver Canucks have taken zone starts to an extreme during the 2011-12 season, with the Sedins and Alex Burrows starting in the offensive end more than 75% of the time, while the Malhotra-Wiese-Lapierre line starts in the defensive zone more than 80% of the time.

Measuring Quality of Competition is less straight-forward: we take the ice time-weighted average of a player's opponents' Corsi number relative to his teammates. The leaders over the last four seasons are a who's who of top defensive players: Nicklas Lidstrom, Willie Mitchell, Dave Bolland, Sammy Pahlsson, Rob Niedermayer, Brent Seabrook, Joel Ward, Jay Bouwmeester, Chris Phillips and Jan Hejda. A primer on the impact of Quality of Competition and Offensive Zone Starts can be found here.

As you can see, the folks at Behind the Net know what they're talking about. They have a good system in place so be sure to check out their statistics page here.

So why use advanced statistics? It gives us a better idea of how a player is actually preforming as opposed to looking at goals and assists. For example, player X on Sydney Crosby's wing could put up great offensive numbers simply because they are on Crosby's wing. Looking at these numbers, you can differentiate what players are helping your team or are just benefiting from being on a line with a certain player.

It can also go beyond that.

Take a look at the website and play around with the numbers. The mainstream media is only now starting to get a feel for how advanced statistics play into hockey but you can be rest assured that the front offices of NHL organizations are more than familiar with these.

Thanks for reading!