The Ongoing Controversy
Analytics has been at the forefront of hockey talk for the past couple of weeks. The conversation started when the Montreal Canadiens failed to renew their analytics guy, Matt Pfeffer, after he argued against trading P.K. Subban for Shea Weber.
Yesterday, a similar incident occurred when the Edmonton Oilers relieved Tyler Dellow of his duties. This may or may not be related to the Taylor Hall–Adam Larsson trade, which caused pandemonium on Twitter.
The Oilers just traded Taylor Hall for a handful of beans. And not even magic beans. Like they reached into a dented can of baked beans.
— Down Goes Brown (@DownGoesBrown) June 29, 2016
Naturally, these moves sparked some serious debate in the hockey community, with some people backing the new mathematical movement and others sticking to the old-school hockey philosophy of mainstream statistics and the eyeball test.
Those of you who know me, follow me on Twitter, or read my articles are aware of my stance on the issue; I am all-in for having analytics play a serious role in hockey. But, before I start defending my stance, there are two things I want to make clear.
1) Analytics is more than corsi and fenwicks
Yes, corsi and fenwicks are advanced stats that factor into analytics, but there is much more than just shot attempts to consider when digging into a player’s on-ice skill.
Statistics like zone-start percentage (ZS%), quality of competition (QualComp), quality of teammates (QualTeam), PDO, zone entries, relative statistics (x-rel), scoring chance differential, the quality of scoring chances, shots-through percentage (SThr%), and game situations all play a role in assessing a player using analytics. These are just some of the metrics used in the process, and there are many more that even I don’t understand yet.
So, while corsi and fewicks are a good introduction to advanced stats, they are just the surface of the craft. Only when you combine them with other data points do you finally get analytics.
"Analytics" is a way of approaching decision making that emphasizes the collection of all available info to make the best decisions possible
— Rhys Jessop (@Thats_Offside) July 19, 2016
2) Analytics isn’t everything
Despite how people tend to perceive the argument, no sane hockey mind will say analytics is 100 percent of the game. Other traits, like leadership abilities, personality, and hockey sense, play a vital part of the game and should be acknowledged when assembling a roster.
The example I like to use here is Dainius Zubrus‘ season with the San Jose Sharks. Zubrus was average in the analytics department and was even a healthy scratch throughout a portion of the 2016 postseason. But, his job wasn’t to be a star player; he was meant to be an example for younger skaters in the bottom six.
He did this by playing a responsible game and being a role model in the locker room and during practice. This helped the development of young players, like Matt Nieto and Chris Tierney, in a campaign that saw San Jose make it all the way to the Stanley Cup Finals.
Though he didn’t even play a full season with the Sharks, his impact will likely be felt for years to come.
So sick of hearing the "numbers aren't everything" schtick. There isn't a single "analytics" guy who thinks numbers are everything.
— Andy Bailey (@AndrewDBailey) April 19, 2016
Analytics will never explain everything because it isn't religion. It will just give you a concrete foundation on which to base decisions.
— Stephen Burtch (@SteveBurtch) May 24, 2016
The Reason for Analytics
Analytics, however, should have a critical part in creating a lineup because it offers so much information in terms of how a player performs and why he performs that way. It can find successes and failures inside someone’s game that the eyeball test may have difficulty picking up.
Let’s consider the following extreme example, which is extremely extreme in order demonstrate the general thought process:
Say there is a rookie (Player A) who comes out and earns 90 points over the course of 82 games in a season. He finishes the year with a plus-68 rating and wins the Calder Trophy.
Given this information, Player A probably sounds like a franchise skater who may become the next Teemu Selanne. But, what if I added these details from the analytics department?
Player A skates on a line with Patrick Kane and Sidney Crosby, and he plays almost exclusively against the opposition’s fourth line. For some odd reason, his club always shoots against the other team’s backup goaltender.
In addition, Player A starts nearly all of his shifts in the offensive zone, and–despite all of these factors–the other team consistently possesses the majority of shot attempts and high-danger scoring chances when he is on the ice. Luckily, his goaltender is the second coming of Dominic Hasek and is able to make the save almost every time, thus giving his team a chance to win.
Suddenly, he doesn’t sound as good anymore. Why?
1) His quality of teammates is extremely high while his quality of competition is incredibly low
2) He plays with a goaltender who has an above-average save percentage and shoots against goaltenders who have below-average save percentages, thus raising his PDO
3) He has heavy offensive zone starts, which means he performs fewer zone entries and likely isn’t trusted in the defensive zone
4) His team gets severely out-chanced while he is on the ice
5) His goaltender faces more shot attempts when he is on the ice
By using basic analytics, we were able to look beyond this player’s point total and plus/minus to find that he is outrageously weak in his own end and has just been extremely lucky in both the offensive and defensive zones throughout his career. These results should not be expected to continue.
All of this was hiding behind a 90-point rookie campaign with a plus-68 rating.
While, again, this is a very extreme and unrealistic example–which I’m using to make my point obvious–the above situation could theoretically happen. So, consider it a proof of concept, if you will.
Obviously, analytics can work the other direction, too, by revealing that a quiet player is actually an elite performer.
Take Marc-Edouard Vlasic, for example. His point totals aren’t flashy, but advanced stats show that he dominates play despite facing the toughest competition in the hardest situations. I’ve made this point multiple times, so if you want an in-depth analysis on how good Vlasic is based on these numbers, go ahead and read this piece.
Public Service Announcement: analytics are important to understanding how good/not so good hockey players are
What a mind blowing concept
— Zibby Fan Account (@DiehardNYRFan) July 18, 2016
Carlyle: "Analytics are a tool, and we have to decide which ones we want to institute as a staff. They're an important part of hockey."
— Anaheim Ducks (@AnaheimDucks) June 14, 2016
Don’t Throw Away Good Data
General managers, coaches, and owners have access to more information than ever before, so it only makes sense to use it. Analytics reveals so much about players and the sport itself that it is borderline foolish to ignore it.
Professionals can make their arguments for size and intimidation all they want, but if mainstream and advanced statistics don’t show results, then obviously the toughness isn’t working. If you say players are afraid to go into the corner against Shea Weber, but he still has a negative plus/minus and negative possession numbers, then guess what?
Players aren’t afraid to go into the corner against Shea Weber.