Second guessing the stats revolution

By Curtis Wolff

There’s a famous brain teaser called the Monty Hall problem in which a game-show contestant is faced with three doors. Behind one door is a prize, while the other two doors hide nothing. The contestant picks a door that he hopes contains a prize. The game-show host then opens one of the other doors, revealing it to be empty. He then allows the contestant to switch his choice of door — should the contestant take the offer?

Statistics say yes. It’s been proven that when the contestant switches doors, he has a two-in-three chance of winning, compared to the original one-in-three. And yet this puzzle caused a furor in the mathematics community, with many established professors remaining skeptical even after computer simulations were run to provide irrefutable evidence. The results just seemed counter-intuitive, even to the brightest minds in the field.

In the minds of sports analytics enthusiasts, hockey has its own Monty Hall problem. While we will likely consider hockey analytics in 2014 quite primitive, the discipline is poised to advance and gain a wide-spread interest. Hockey analytics have struck a chord with a niche group of fans and bloggers — a group which is becoming increasingly frustrated by the hockey establishment’s reluctance to embrace their research.

Take, for example, Calgary’s own Brian Burke. The Flames’s president of hockey operations was recently invited to the 2014 Sloan Sports Analytics Conference as a panelist. He was largely dismissive towards current advanced hockey statistics, much to the dismay of the enthusiastic stats nerds assembled at the Mecca of sports analytics.

Burke made similar comments during his January visit to the University of Calgary, after fielding a question from an enthusiastic audience member about advanced analytics as the “future of scouting.”

“To me statistics are like a lamppost. They are useful for support, but not for illumination,” said Burke, who says that he personally reads dozens of analytics papers from grad students every year and hasn’t yet been satisfied.

“I have a very healthy skepticism about it because this isn’t baseball,” Burke continued. “These are random events that don’t repeat. Baseball is perfect for it.”

Burke has a point. Hockey is difficult to capture with statistics compared to numbers-darling sports like baseball, which consists of highly repetitive events which are easily charted.

Many advanced hockey statistics put forth a good effort, but leave much to be desired. For example, Corsi is a popular advanced statistic that measures the attempted shots for and against a team while a certain player is on the ice. However Corsi is then used to gauge a player’s possession ability, even though all it can literally measure is shots directed at the net by the entire team while a particular player is on the ice. While obviously skilled players tend to have high Corsi ratings, and blatantly bad players usually have poor ones, there’s still a nagging gap between what Corsi measures and how it is interpreted.

Some stats reveal tendencies that are fascinating for curious fans, but are glaringly obvious for NHL coaches. Look at statistics which measure zone start and quality of competition. You don’t need to tell Flames head coach Bob Hartley that he gives Dennis Wideman mostly offensive zone starts or that he lets Mark Giordano play some of the toughest defensive minutes in the league — he’s the one who makes those calls.

There’s also a fundamental divide between the thought process of a coach and statistics. Statistics tell us a story about the past, but a coach’s job is to get the best out of his players today and in the future. While coaches can use statistics to help chart the future, clashes between a coach trying to manage a group of human beings and a statician’s numbers on a graph are inevitable.

This is not to say that these statistics aren’t useful. They are simply lacking the gravity that staticians claim they have. There are very few tremendous insights to be gained from these stats that couldn’t be revealed by watching the games. Statistics can fill the gaps when a team’s scouts can’t watch every game, but ultimately fail to provide new information, especially to NHL GMs and scouts who watch hockey for a living.

Mocking those like Burke and labeling them as dinosaurs is easy. They are the old guard who are ignoring the young whippersnappers and their progressive ways. These analytics are interesting numbers for fans to toy with and even for coaches and managers to use in certain situations. But in reality these stats are not ready to replace human scouting or play a major role in how NHL teams make decisions. The game is too fast with too many variables, and the math just can’t keep up.