Four things to remember about hockey analytics

Statistical analysis in sports always seems to be met with a lot of intrigue or backlash depending on who you ask. There are some who embrace a more analytical approach to the game while others find this stuff to be hogwash and a waste of time. Some of the detractors go as far to say that they have no place in sports and no one cares about them.

While stats are likely an afterthought to most players, they do have their place in the game and they have come a long way over the last decade or so. Hockey is still behind the curve compared to baseball, basketball, soccer and football, but it's not completley irrelevant. There are some teams & coaches who use analytics when making decisions, so this isn't something that only appeals to nerds in their parent's basements. 

Still, I feel that statistics in sports are very polarizing because there is a belief that it's impossible to quantify everything that happens in a hockey game. It's true that the game moves very fast and it's difficult to catch every event, but it's silly to disregard analytics simply because of that. I've always thought that statistics are a way to enhance one's analysis of the game because relying on the eye-test alone can be just as deceiving. The human mind is very selective and one is more likely to remember the noteworthy events of a game rather than everything that happened.

For instance, a defenseman might create a brutal turnover that leads to a goal against and this might cause a lot of spectators to believe that this player is terrible and cringe every time he has the puck in the future. They might forget that he also made a lot of good plays to exit the zone, deny a scoring chance or keep a puck in at the blue line to help set up a scoring chance or a goal for his team. Stuff like that is typically missed by the casual fan and taking a closer look at the game through analytics can help catch events such as this. It may not be everyone's cup of tea and that's fine, but they are worth taking into consideration.

However, as helpful as analytics can be, they can also be really overwhelming to someone who is completely new to them. That's mostly because there is a lot to sports analytics and it definitely takes some practice and a lot of understanding for those to use them correctly. Stats are often misused and misinterpreted because of this and it seems to be growing now with more people discovering sites like Behind the Net & Hockey Analysis. It's great that this is catching on, but I still think a lot of people make a lot of assumptions about certain hockey statistics and don't put a lot of thought into them other than what they are at face value. 

As someone who uses analytics in basically every blog post, I've been planning on writing a guide/overview of hockey stats and how to use them for quite some time now. Unfortunately, time constraints has prevented me from doing so and a lot of other great bloggers have written their own guides on this subject. Each of those are well-done and worth reading, so I won't bother copying them. Instead, what I'm going to do is give newcomers a few tips on how to use these "advanced stats" to make the learning process somewhat easier. Stats only mean so much at face value, but how you interpret and utilize them can really mean a lot.

Why do I use analytics?

Before we get started, I'll talk about why I use this approach over relying on just the eye-test. For me, I've always had a bit of an "obsession" with wanting to learn more about a subject and dig deep for information. Whether it was geography, music or sports, I always liked to learn as much as I can about something I was interested in and it eventually led me getting into in-depth sports statistics over the last few years. The analytical approach has always been more appealing to me because there is a lot you can learn about a team or a player through this than by just watching the games. Watching the game is important, it just isn't enough if you want to learn as much as you can.

With hockey, puck-possession metrics like Corsi/Fenwick were confusing at first (mostly because of the silly names) but they really make a lot of sense once you learn about them. Corsi/Fenwick is the bread & butter of most hockey statistical work and all they are is the shot attempt differential of a team or a player during five-on-five play. Once you get past the names, it's fairly simple to understand and it's not a terribly foreign approach. Teams who can possess the puck more create more offense, spend less time in their own zone and are more likely to score more than the opposing team. This is basically what stats like Corsi, Fenwick and scoring chances try to capture. However, possession isn't everything and this provides a nice transition into the meat of this post.

Things to Remember About Hockey Statistics/Analytics

1. Nothing is absolute.

This is one of the things that any stats person should tell you but for some reason, a lot of people mistake statistical analysis to be "end be all." They will dismiss a stat because a player they perceive as good has a terrible Corsi percentage or vice versa. While it is true that a lot of players who spend a long enough time in their own zone to be outshot by a considerable margin are legitimately awful, it doesn't always work this way. Possession is just one part of a game and while a player's strength may not be in that area, there might be other areas that he excels in.

While it has been shown that being a strong puck possession team increases your chances of making the playoffs, it isn't something to base all of your analysis on, especially if you are looking at only one player. Linemates can play such a big role in controlling territorial play and it's not always simple to find out who is driving the bus, although looking at how a player performed with and without a certain teammate on the ice can help this a lot. Still, possession is only one part of the game, albeit a very important one, and there are other factors worth taking into consideration when using analytics to judge a player's performance or value. 

Because of this, some might write off Corsi/possession as a "flawed" stat and dismiss it all together. In my opinion, this just comes across as lazy and a way of saying "this is different from my way of thinking so it's wrong." If that's your way of thinking then that's fine because Corsi isn't perfect, but I don't like the idea of dismissing a stat just because it doesn't tell the whole story or it says something you disagree with. The thing is, no stat can tell you the entire story and that includes ones that are regularly used by the mainstream audience (goals, points, plus-minus, hits, shots on goal, blocked shots, etc.). Corsi & possession stats are just away to expand on these stats by looking at more events and giving viewers a bigger picture of what happened during a game.

That's what the goal of "advanced stats" are, to help give you a better picture of what happened during the game and possibly catch a few things your eyes may have missed. Unfortunately, with only so much information being available through the NHL's web site, there is a limit to how much some of these stats can tell you, which is why a lot of bloggers have taken it upon themselves to track events such as scoring chances, zone entries & zone exits. It takes a lot of hard work to do this type of analysis, but the information gained through this is very, very rewarding for those looking to dive deeper into hockey analytics.

Right now, there isn't one stat that tells you everything there is to know about a player's value and there probably never will be because of how many different elements there are in hockey. However, this doesn't mean that the stats we use now are worthless because possession is a big part of the game and there is a lot you can learn about certain teams or players through this. It just won't tell you everything and there are a lot of other things you need to look at. You have to be willing to take the time and do the work to learn more about these stats instead of just taking them for what they are at face value. 

2. Context is important

This is such a vital part of doing any type of analysis involving possession metrics a player's linemates, opponents and territorial position is going to have a pretty big impact on how many shots he is on the ice for. Take a player like Brandon Sutter for instance. During his time with the Hurricanes, he was the coaches "matchup guy" and was always sent out against other team's top lines. He also started a majority of his shifts in the defensive zone, which gave him a slight territorial disadvantage compared to players like Jeff Skinner, who were deployed more in the offensive zone. As a result, the Hurricanes were usually outshot when Sutter was on the ice. 

Now, this doesn't meant it's not impossible for a player to win the possession battle with tough assignments because players like Jeff Halpern, Boyd Gordon & Chris Kelly managed to do it this year, but these players have a bigger workload than guys with cushier minutes. So when you say to yourself "no wonder Mikhail Grabovski got bought out, his Corsi rating is awful," remember to note that he took 36.7% of his 5v5 draws in the offensive zone and played a lot of his minutes against opposing team's top lines. Thus, he was playing in a very difficult role and it was going to be difficult for him to create offense because he was at a territorial disadvantage.

Likewise, a player with easy assignments can post unreal possession & scoring numbers because he is starting the majority of his shifts in the shadow of the opposing goaltender. The Sedin twins, Brendan Gallagher and Evgeni Malkin being examples of this. However, getting a lot of shifts in the offensive zone doesn't mean a player will outshoot his opposition because there are a lot who struggle to win the possession battle despite soft minutes, so their numbers would be very unimpressive. Those who can't keep play in the offensive zone while getting favorable territorial assignments are generally looked down upon in the stat community.

3. Regression, luck and shooting percentage plays a role

This is probably the most difficult thing to get people on board with, but luck plays a factor in hockey especially when it comes to goal-scoring. There is no denying that it takes a lot of skill to score and I'm a believer in shot quality, but take a moment to think about some of the goals you've seen. How many of them came off a weird bounce, a deflection, a bad play by the goaltender or some sort of fluky occurrence? Probably more than you think. Sometimes a player can do everything to beat a goaltender but end up hitting the goal post or go off the side of the net. There also a lot of times where players can't get to rebounds in time or have one cleared away at the last second.

Hockey is a fast-paced game and fluky bounces can happen relatively often, many of which can end up in the back of the net and they don't always even out over the course of a season. This was especially true for last year with the shortened schedule. Fact is, in a game where chance can decide the outcome, strange things can happen and sometimes a team can dominate their opponent but still lose. They might not be able to buy a goal while having a couple of mistakes end up in the back of their own net. Those kinds of things tend to happen and while they are frustrating beyond belief, there isn't much you can do about it other than go with a similar gameplan next time and hope for a better result.

This is often a huge factor in a playoff series where one hot streak or elite goaltending can be the deciding factor for a team in a seven-game series, but it definitely can be a huge factor in the regular season, as well. We saw that with the Hurricanes this year. The team's underlying numbers improved greatly from last season when the team couldn't buy a goal and got sub-par netminding for the last 20 games. In a full-82 game season, the Hurricanes may have been able to climb out of this rut but with a condensed schedule, that's more than enough to sink your season. The New Jersey Devils suffered a similar fate, as they had the best 5v5 shot differential in the league but could not find the back of the net to save their lives. Regression played a role in the resurgence of teams like Los Angeles & St. Louis during the 2011-12 season, so it's possible that the same may have happened with Carolina & New Jersey in a full-season. That isn't guaranteed, but the numbers suggest that they were nowhere near as bad as they finished this year.

Regression also has an impact at the individual level when it comes to a player's goal scoring. Let's say a player scores 30 goals one year while recording 2.6 shots per game. He improves his shot rate to 3.3 the next year and scores 25 goals, then the following year he improves his shot rate to 3.7 but scores only 18 goals. Going by just goal scoring alone, you could say that this player is declining but he is actually producing more offense without getting much of a reward for it. So you could say that this player may have had a lot of things go his way in that 30 goal season, while getting terribly unlucky in the 18-goal season. Since his shot rate steadily improved over the years, it's fair to claim that he may have not gotten a ton of bounces to go in his favor that year because he was doing his job in terms of producing offense. Luck can factor into goal scoring and sometimes it can make a player worse than he appears.

With that being said, it's important to look at a player's career shooting percentage in mind when seeing how "lucky" or "unlucky" he got in a year. For instance, Ryan Getzlaf was terribly unlucky in 2011-12 season when he scored only 11 goals. We can say this because Getzlaf is a career 11.9% shooter and we would expect him to score on more than 11 of 159 shots if we were to replay that season. The 8.8 shooting percentage Jordan Staal posted this year is also lower than his career average and it's reasonable to expect him to rebound next year. On the other hand, we probably won't see as big of a rebound from David Clarkson, whose 8.3 shooting percentage last year was very closer to his career average, which means it is about what most expect from him. This sort of relates back to the point I made about context needing to be taken into account.

4. Analytics are still growing

Possession stats are all the rage right now, but they are just the tip of the iceberg because there is really a lot you can dive into in the world of hockey metrics. Unfortunately, most of them are not made available to the general public, so this is what we currently have to work with. Bloggers like myself, Josh Smolow of Lighthouse Hockey, Eric Tulsky of Broad Street Hockey, Oliver Bouchard of En Attendant Les Nordiques, Josh Lile of Defending Big D and Jonathan Willis of Oilers Nation have taken the initiative to track things like scoring chances, zone entries, zone exits and a lot of other information during games to help enhance our analysis and better understand the ins and outs of our respective teams.

You may remember me posting zone entries from the Hurricanes this past year to illustrate their performance in the neutral zone and tracking scoring chances for the last two years. My hope is to continue and expand on these studies in future seasons so that we can have as much information available as possible. It will take a grassroots movement by a lot of bloggers for this to be available on a league wide basis, unfortunately but more information available now than there ever has before and it's good to see people embracing this. 

However, there is still a bit of a hostile attitude towards the world of hockey analytics when I feel there shouldn't be. I understand that stats aren't everyone's thing and they go against a lot of sports journalists obsession with narratives, but I've always been of the mindset that if you have the opportunity to learn more about something, then you should take full advantage of it. This is why I've embraced the use of analytics in hockey and other sports and it's really enhanced my enjoyment of the game. Stats are only useful when they're interpreted correctly, though, so it's important to keep context in mind when demonstrating them.

Quantcast