Hockey And Big Data

Editor’s note: Nick Rojas is a business consultant and writer who lives in Los Angeles and Chicago. He has consulted small and medium-sized enterprises for over twenty years. John Siegel is a writer and hockey enthusiast.

Hockey is a naturally aggressive sport, and the casual fan has learned to associate it with violence — the kind that makes the daily sports highlight segment on the news.

But like other sports, there are subtle nuances to the game that are lost in between the bare-knuckle brawls and bone-jarring hits. Hockey is the ultimate team sport in the regard that players are more than willing to sacrifice notoriety — and the physical well-being — for the betterment of the organization.

Whether it’s Mark Recchi’s kidney stone surgery (in addition to broken ribs) on the morning of game 7 of the Stanley Cup Final, or Sebastien Courcelles having his face slashed open with a skate, there are countless instances — both obvious and hidden — that show how the players exhibit a toughness that’s rarely matched in an athletic setting.

The business of sports, however, can’t quantify the toughness of its players. This is where hockey finds common ground among most other professional sports. While hockey players tend to their wounds and clean up the blood, the front office will likely have another team working within a discipline that can also be described as unforgiving, beautiful and even brutal, but for very different reasons. At least that’s how scientists sometimes describe math.

What’s at work is an applied science, and as usual, science rarely gets it right the first time. In this case, it’s also about getting a lot of help from people in big data and analytics from overlapping fields. What’s really new about hockey and analytics can be found at the very beginning and end of the process, namely what data should be collected and in what form the results should be expressed.

As with baseball and Earnshaw Cook, players (that is, players in the game of hockey analytics) are emerging with ideas about effectively converting raw data into decisions for a winning team.

So what are teams doing to actually produce results? Although the objective is simple, strategies for getting there can vary from team to team. Kyle Dubas, the young assistant GM of the Maple Leafs, created a hockey “research and development” team. It includes a chemical engineer and a mathematician. Their only task is to  apply science and engineering to hockey statistics, resulting in more wins for the team.

In addition to in-house talent, teams will be working with academic institutions and companies that specialize in big data and analytics. One such company is SAS Analytics, which does about $3 billion in business annually. They use big data to help banks, telecommunications companies and even governments. Although most of their work has been done outside of sports, the datasets will often be treated in similar ways if compared to other projects.

The convergence of math, computer science, information technology and (as always) the Internet of Things (IoT) will also drive the end result. Whether it’s a large corporation or a fan, both professional and amateur analysts will likely be building on existing ideas to varying degrees. HARO, HARD and HART ratings are examples of simple calculations currently in use, where:

HARO = Hockey Analysis Rating Offense
HARD = Hockey Analysis Rating Defense
HART = Hockey Analysis Rating Total

The foundation of good results is good data. People like Corsi and Fenwick (for example) have made a name for themselves by being among the first to develop some usable shot-based analytics. In simplest terms, the analysis amounts to a tally of shots along with the outcome of each shot. Compared to some of the newer ideas, it’s not especially sophisticated, but it’s a good start.

When applied, it looks something like this:

A player (player “A”) is on the ice for 10 shots during a game. During that time interval, the opposing team takes three shots:

Corsi For (CF) = 10
Corsi Against (CA) = 3
Player “A” would be a +7 Corsi (10 – 3 = 7) for the game

It’s important to note that typically, shots in the game of hockey are only counted in the box score if they make their way to the goalie, who either saves the shot or lets in a goal. Corsi and Fenwick take into account all shots direct towards the net, which also includes shots that miss their mark or are blocked by an opponent.

Let’s say that during the time that player “A” was on the ice, 10 shots were made, two were blocked by players on the opposing team. The opposing team had three shots during that same time interval, but one was blocked. Fenwick excludes blocked shots from the formula, so the numbers for player “A” would look like this:

Fenwick For (FF) = 8 (10-2)
Fenwick Against (FA) = 2 (3-1)
Player “A” would be a +6 Fenwick (8 – 2 = 6) for the game.

To make the data easier to use, statisticians express the results as a percentage. In this case:

CF% (Corsi For Percentage) = CF/(CF+CA) = 76.9%
FF% (Fenwick For Percentage) = FF/(FF+FA) = 80%

It’s important to appreciate the volume of information. Player data can amount to millions of data points. The quest for accurate data and powerful analytics in sports will change shots into goals, and goals into cash. At least, thats what the teams are banking on. In addition, predictive analytics may also be used to anticipate injuries.

There’s evidence to support a correlation between winning teams and the numbers they produce during the season. For example, The Los Angeles Kings were the 2012 and 2014 Stanley Cup winners. Their CF% and FF% are as follows:

Los Angeles Kings Team Stats CF% and FF%

Year

CF% (rank)

FF% (rank)

2012

54.8 (2)

53.6 (2)

2014

56.8 (1)

56.1 (1)

Source: hockeyanalysis.com

A significant roadblock to the widespread implementation of “fancy stats” around the NHL lies in the way that the game has been taught — and subsequently, played — as its popularity grew over the course of the 20th century.

Frankly, the machismo of the game stands firmly entrenched in front of the evolution of the methods by which it is played. Enforcers, long used to protect smaller, more talented players, are slowly but surely being phased out — as is the act of fighting — so that more talented players can take their roster spot. In fact, fighting has been on the decline since the late 1980s.

Active debates among peers can already be found on the Internet. What’s interesting is the role that fans may play in the peer review process (whether team owners want to acknowledge them or not). Fan analytics is quite competitive, and deserving of some credit for evolving what had been a stagnant approach to the progression of the game.

There is a serious effort among fans with the aptitude and will to understand it. At times, the word “effort” is a bit rosey when one considers some of the lively exchanges regarding the benefits of traditional analytics, so-called “advanced analytics,” and no analytics. Even the classification of methods is fair game.

With all the attention given to this new avenue for getting ahead, it will be interesting to see how it’s incorporated into various aspects of the sport, from the top-secret “teams within a team” to the fan experience, both at home and near the ice.