Analytics – So What?

Analytics. The buzzword in the hockey community these days. Big Data is quickly moving in to the hockey world and changing the way we look at the game. There is no doubt that this will change hockey at a number of levels, but ultimately it’s about winning hockey games.


Goals Scored in the 2015 NCAA Tournament: Regionals

This weekend, there were 12 games played in the NCAA Division I Men’s Hockey Tournament. Those 12 games resulted in four regional champions that are heading to Boston for the Frozen Four.

In those 12 games, there were a grand total of 66 goals scored. I took a closer look at those 66 goals, where they came from and if there were any trends that could be noticed. I looked at a few different categories on each goal – strength, time in zone prior to goal, zone entry type (carry, dump, faceoff), goal type (rush, in zone possession, forecheck, faceoff, empty net), clearing attempts prior to goal, turnovers prior to goal, and lost battles prior to goal. Here’s a brief summary of the numbers:

  • Of the 66 goals:
    • 9 were empty net (13.64%)
    • 14 were on the Power Play (21.21%)
    • 3 were Shorthanded (4.54%)
    • 40 were even strength (60.61%)

Let’s look closer at the 40 even strength goals

  • 31 of the 40 (77.5%) came after the puck was carried into the offensive zone
  • 20 came off the rush
    • Average length of possession in zone was 5.05 seconds from entry to goal scored
  • 14 came from offensive zone possession
    • Average time in zone was 22.07 seconds from entry to goal scored
    • 13 of 14 came from at least one lost battle by the defending team (92.86%)
    • 9 of 14 came after the defensive team had an opportunity to clear (64.29%)
    • 8 of 14 came after a change in possession in zone (57.14%)
  • 1 came off of a faceoff
  • 4 were the result of good forechecks

Taking a closer look at the goaltending (57 goals allowed):

  • 15 of the 57 (26.32%) goals beat the goaltenders clean (goalie had time to set on the shot)
  • 18 of the 57 (31.58%) came immediately following a pass
  • 12 of the 57 (21.05%) were scored on a rebound
  • 12 of the 57 involved traffic at the net – either a tip (3 – 5.26%) or a screen (9 – 15.79%)


  • I had suspected prior to doing this research that a good majority of goals were scored after a failed clear. While it is a very small sample size, about 65% of goals scored in the offensive zone come after a failed clear.
  • I was surprised with the high number of rush goals – having half of the goals scored at even strength be off the rush is a surprisingly high number.
  • I am not surprised that the number of goals after a lost battle is very high. Often teams that maintain possession do so as a result of winning puck races and 1v1 battles – the longer you possess the puck, the more fatigue sets in and the higher your chances of scoring.
  • Turnovers in the defensive zone are especially damning as well – 57% of goals scored off off OZ possession come after a turnover.
  • The number of goals scored that beat the goaltenders clean was surprising. Over 25% were shots that beat a goaltender that was set on the shot. More on par with expectations was the number of goals after a pass and off of rebounds.

It is a small sample size, but it is very interesting to look at and see how goals are scored in the biggest games in Division I hockey. Bottom line – execute your clears, don’t turn the puck over and limit possession time in the offensive zone for your opponent…things that all good coaches preach on a regular basis.

Update: Here is the data set that I used in a pdf form:

Special Teams Analytics

I’m fascinated by the use of high level statistics in the NHL. I think it’s a great way to look at the game from a different angle and challenge previously held beliefs that are common in the game.

One example is special teams. Everyone in the hockey world looks at PP and PK% as an indicator of their success. While it is a measure of past success, it is not an accurate indicator of future success. Fear the Fin, a blog about the Sharks did a great job breaking down special teams from an statistics point of view. A fascinating look at the different factors that affect a power play (other than tactics and personnel) or penalty kill and how a team’s success or failure in those situations affects their place in the standings over the course of a season. Check out the full article here. The biggest takeaways for me are in the summary:

Summary Points

  • NHL teams spend a substantial portion of games on special teams. 5v4 time alone accounts for 10-20% of game time, and thus needs to be analyzed to further our prediction models.
  • Although we don’t have enough years of NHL RTSS (5 years and counting) data to conclude with statistical significance, it appears that Fenwick For/60, with misses and blocked shots adjusted for scorer bias, is the best predictor of power-play success (GF/60), and well correlated with winning (Pts/game).
  • The penalty-kill picture is less clear, likely the result of heavier regression to the mean. Presumably the heavier regression is because PK units spend much more time without the puck. Both Sv% (which is likely goaltender driven), and Corsi differential/60 (Corsi For – Corsi Against per 60) are predictive of future penalty kill success (GA/60), and winning (pts/game), but less powerful than the predictors of power-play success.
  • Shooting percentage on the power-play is negligible, regressing heavily to the mean, and shows at best very modest correlations with PP success and winning. Even if we attempt to attenuate Sh% (ie. control for how much it regresses to the mean), it still under-performs other more significant metrics like FF/60.
  • If we focus on which stat in theory has the strongest association with PP success (GF/60), we see that Fenwick For per 60 and Shots For per 60 are virtually indistinguishable. On the PK, however, Sv% becomes the strongest stat.

Looking a little closer at PP and PK is the Flyers blog Broad Street Hockey. In two articles, they dig into PP and PK and how you can look at future success for each of them. Their bottom line conclusion? PP success is indicated by shot rate while PK success is a little simpler, with just conversion rate. You can check out the PP article here: and the PK article here:

What drove me to read these articles? A very interesting piece on TSN’s Analytics Blog by Travis Yost going into the details of Tampa Bay’s success at even strength and their relative weakness on special teams. Tampa has a power play that is clicking around 20% of the time and a PK that is successful about 85% of the time. Pretty good, no? Well looking deeper, they actually have a shot differential of -35.7 per 60 minutes of special teams play. What the article argues is that their current success rates are unsustainable based upon their shot differential. It will be interesting to see what happens moving forward with Tampa’s special teams units. You can find that article here:

My takeaways from these articles and insights into statistics? Well it also passes the “eye test” with hockey. Want to have a successful power play? It comes from shots. Shots come from faceoff wins and puck retrievals in the offensive zone. Want to have a successful penalty kill? A good goaltender goes a long way – as does limiting the quality of shots against. Finally, you can look at analytics to see if your success on special teams is due to random events or methodical success through statistical trends.

I’ll be honest, as a history major in college I never took a stats class and I couldn’t tell you what the actual equations and statistical work means. However, I trust that you need that data to prove your points and that the greater analytics community would immediately invalidate any argument that doesn’t pass the statistical data test. I try not to get too lost in the math but instead interpret the numbers and determine what they mean and what the effect on the game might be.

Scoring Chances

There are many ways to measure performance in a game – eye test, scoresheet, standard metrics, advanced metrics, etc.

To me, the best indicator of overall performance is Scoring Chances. We define a scoring chance as any shot on net from within the “scoring area” (inside the dots and below the tops of the circles).Statistically speaking, NHL goalies have a .855 save percentage inside the scoring area, and a .958 save percentage outside of it (reference: Any shot from within the “scoring area” has a much greater chance of beating the goaltender.

Scoring Area

Why do I consider this the best indicator of performance? Critical moments. Scoring chances are the critical moments in any game that dictate the outcome. There are thousands of innocuous plays in every hockey game, but only about 20-40 that qualify as a scoring chance. How you perform in these instances says a lot about how you played the game.

How do we measure scoring chances? After every game, I go through and watch every shot for and against. The shots that are released from within the scoring area are recorded as scoring chances. I then record responsibility for each chance – for and against. This is the subjective part of scoring chances, sometimes it is hard to say who is more responsible for a certain play. We assign primary and secondary responsibility for every play – the difference between primary and secondary is sometimes marginal, but there are almost always at least two players who could have changed the outcome of the play. I don’t look at primary vs secondary very often, mostly overall scoring chance +/- (your involvement in chances for minus your involvement in chances against).

We also measure a stat called absolute scoring chances. Absolute scoring chances measures the number of chances for and against while you were on the ice, regardless of involvement in the play. This indicates if a player, line, d pair is more of a positive or negative influence on the game overall. I also feel this is a good indicator of matchups – if you were playing against an opponents top line and were even or better in absolute scoring chances, that is a good game.

Here is a look at our stat sheet from after a typical game:


Total +/- indicates a player’s involvement (primary or secondary) in scoring chances. Absolute +/- is their on ice presence for any even strength scoring chance. In this game, two players had poor performances (-4 and -5 total) while a few (the +2s total and +3 and +4 absolute) had good performances.

While there is always a big picture evaluation, scoring chances gives you a snapshot of who was involved and who influenced the game in a positive or a negative manner. Teaching through scoring chances gives you an opportunity to improve your performance in the critical moments that define a hockey game.

Goaltender Quality Starts

This weekend I spent a lot of time in the car on a recruiting road trip with a good friend. During our hours in the car, I was introduced to the concept of Goaltender Quality Starts. The more I thought about it, the more I liked the idea as a baseline measurement of a goaltender’s performance.

A Quality Start is defined as any game where a goaltender has a Save Percentage higher than .912 (league average). Quality Starts can also be awarded in games where the goaltender has a save percentage between .885 and .912 and gives up less than three goals. This controls for games where goaltenders see 20 shots and give up 2 goals, for example. These numbers are based upon winning percentages – when a goaltender has a .885 or higher SV%, teams have at least a .500 winning percentage.

While these numbers don’t speak to the nuances of a goaltenders game (how he controlled the puck, his skating, save selection, etc) they do put a measurement on the end result and allow coaches and statisticians to quantify performance. The stat is also very illuminating when it comes to winning and losing. Our record last year when our goaltenders had a quality start was 7-1-1 (.833 winning %). When we did not receive a quality start, we were 0-14-2 (.067 winning %).

I like the use of a quality start metric as a benchmark to measure performance – did a goaltender give you a chance to win?

A full explanation and the details of the stat can be found on the hockey prospectus blog here (full credit to Robert Vollman for his work and writing):

The Importance of the First 20 Minutes

I recently completed a study on the 2011-2012 NCAA hockey season and the situational records of every team. I looked at the records for every team in Division 1 and their wins and losses in home/away games, after each period, after the first goal, and in games of certain scoring margins. I also broke it down by conference and by teams that made the NCAA Tournament. All of these yielded interesting results, but two in particular jumped out at me.

The first number that caught my eye was team records after the first goal. Everyone in hockey wants to score the first goal, but I know that I was unaware as to the impact of that goal on the result of the hockey game. In 2011-2012, teams that scored first won 66.94% of their games. If you just look at teams that made the NCAA Tournament, that number jumps to 78.02%.

The second (and much more striking) number is winning percentage after the first period. Teams that had a lead after the first period won 77.09% of the time. Once again, this number jumps among teams in the Tournament, up to 84.75%.

An old adage in hockey is to “Use statistics like a drunkard uses a lamp post – for support, not illumination.” I do not disagree with this statement and in this case these numbers support the importance of scoring first and having a lead after the first period. Naturally, every team wants to score first and maintain a lead, but I was unaware as to the amount that this influenced a game. Now that I am aware, the process moves to how am I going to use these numbers to make myself a better coach and my team more successful.

Scoring (Chances)

The game of hockey is a fast paced flow sport that demands many different actions and reactions for each player. These events can be seen and understood in many different ways and how success or failure is measured can, at times, be subjective. One of the more cut and dry areas of hockey analytics and statistics is scoring chances. Every team that plays at a high level looks at the impact of scoring chances on a game. It is safe to assume that most teams always attempt to out-chance the opponent.

What is a scoring chance? There is no universal definition, but it is a stat that is universally tracked. At UMass, we define a scoring chance as a shot from Grade-A (the area in between the faceoff dots, from the top of the circles down to the goal line) that hits the net. While this eliminates shots that hit the post and pucks that roll untouched through the crease (among others), it is the most objective way to look at scoring chances. We also look at scoring opportunities which include plays that would have been a chance if the puck was on net.

Why are scoring chances important? In hockey, the object is to score more goals than the opponent. While there are always outliers, it is safe to say that teams score more goals from the scoring area pictured than from the perimeter of the ice sheet. As a coach, I want to know how many scoring chances my team generated and how many we surrendered. I also want to know what our shooting percentage is on scoring chances and what our goalies’ save percentage is on scoring chances against. These are numbers that I can use to quantify our performance in the most important area of the ice, both offensively and defensively.

Scoring chances and scoring chance involvement can be tracked on an individual level as well. While this begins to enter the more subjective range, it is a way to see how a player is getting involved both offensively and defensively. Defensemen will often be skewed towards the negative, while Forwards will skew towards the positive, as this is the nature of their position. For this reason, individual chance performance should be measured against a few different numbers, including team averages, linemate averages, and averages per shift (40 seconds). This allows players to be fairly evaluated against the overall performance of their teammates and more specifically to the players they play with on a frequent basis.

I certainly understand and appreciate that there is much more to hockey than scoring chances, however I feel that looking at and tracking scoring chances is the most effective way to get an overall view of a team’s performance. It measures the ability of the player and the team to perform when it matters most.