Analytics are a very useful way of evaluating players and teams. A large problem with analytics is it seems very complicated. The common misconception among the general fan population is that one requires mathematical ability to comprehend and draw conclusions from a set of data. This is completely untrue. Advanced stats are no different than run of the mill box stats. The only people who truly need any form of mathematical training are the ones who develop these stats.
Analytics are not the enemy of the eye test. In fact most stats that are logged into databases are tracked by hand, by people who watch the games. Advanced stats are simply an organized version of the eye test which allow us to view the results of a player. I don’t care how a player looks while he is playing, as long as he is playing well. Even for those who don’t fully buy into an analytical approach, using advanced stats as a way of confirming or denying existing opinions or observations can help someone become a more knowledgeable fan. The way I see it, more information can’t be harmful.
I’ve decided to write this with the intention of informing those who might have been otherwise uneducated. I’ll do my best to explain some of the basic stats, as well as provide a list of useful resources for those interested in further pursuing analytics.
Corsi (C): This is simply a measure of shot attempts. Instead of counting shots which hit the goalie, it counts shots that were blocked on the way to the net or missed. Corsi is probably the most important advanced stat, because it’s the best at predicting success without a large sample size.
Fenwick (F): This is another shot differential stat, almost the same as Corsi. The only difference is that is doesn’t count blocked shots.
Corsi for percentage (CF%)/Fenwick for percentage (FF%): This is the differential of shot attempts. Anything over 50% means the team is controlling the majority of shot attempts, which is obviously good. Anything below 50% means that on average, the team’s opponent attempts more shots. Typically teams are between 48%-52%. Typically when people refer to a team’s possession, they are referencing their CF%.
Expected goals (xG): Over large sample sizes, expected goals is the best at predicting success. Basically this means, based on how many times a player/team attempts shots and where those shots were taken from, we can estimate how many goals this team/player should be expected to have scored. If the actual amount of goals this player scored was significantly above or below their expected goals, then we could determine this player was probably lucky or unlucky.
PDO: This is essentially the statistical quantification of luck. The underlying theory being that players have little to no control over their on-ice shooting percentage and on-ice save percentage. PDO is calculated by adding the on-ice shooting percentage and save percentage of a player or team while they are on the ice. Anything above 100, means this team/player is lucky, anything below 100 means a team/player was unlucky. This stat can be useful, for predicting success as well. For example, the 2013-2014 Colorado Avalanche went from being the 27th ranked team to the 3rd in just one season. However, their PDO was ridiculously high at 102. Advanced stats people predicted them to fall back to earth the next season, and they haven’t been back to the playoffs since.
Goals for percentage (GF%): This stat refers to the percentage of goals scored in a team’s favour when a specific player is on the ice, or the team as a whole. Same with CF% and FF%, you want to be over 50% because this means you are outscoring your opponents. This stat is not considered to be an accurate reflection of an individual player’s talent, for the same reasons that +/- is not very good. It can be too heavily reliant on save percentages or the quality of a player’s teammates.
Quality of Competition (QoC)/Quality of Teammates (QoT): One of the common misconceptions in hockey is that different players face drastically different competition. For example, many believe that Shea Weber and Roman Josi suffered from poor possession numbers because they faced top competition. In reality, the difference between those facing top competition and those facing easier competition is quite small. This is because player take such short, frequent shifts that they are bound to face many different types of players. Basically nobody plays exclusively against Sidney Crosby or against Tanner Glass. However, players do experience a large fluctuation in their quality of teammates which can impact their stats quite heavily. If somebody is playing on a fourth line, they are likely playing with other fourth line quality players, and maybe 3rd pairing defensemen.
This is an excellent visual done by Sean Tierney (@SeanTierneyTss on Twitter) which demonstrates how quality of teammates has a much wider gap than quality of competition. Both QoC and QoT are measured using CF%, xGF%, Time on ice (TOI) or sometimes GF%. None of these are perfect metrics for determining the true quality of players, but are presently the best ways possible.
Primary Points (P1): This is not really an advanced stat, but nevertheless I receive questions on what exactly a primary point is and what the true value in using it is. A primary point is quite simply a goal or a first assist, meaning that secondary assists don’t count. Primary points are better than points because they are better reflections of a player’s individual talent. They aren’t as reliant on teammates. Some believe that secondary assists have no value whatsoever.
Points per 60 (P/60): This is the amount of points a player scores per 60 minutes of ice time. This is better than using points per game because often times points per game can be inflated or deflated by how much or little a player plays in a game. If player A is playing 20 minutes a night and scores a point per game, and player B plays 10 minutes per game and averages half a point per game, it’s not fair to say player A is better offensively even though his point per game is twice as high.
Relative stats (Rel.): This is simply a measure of any given stat compared to their team average. If a player has a high relative CF% means they are outperforming their team in terms of shot attempts, low means they underperform their team. Relative stats can be applied to a variety of stats such as Corsi, Fenwick and goal totals.
I believe part of the reason many don’t use analytics is because you can’t find them on the NHL website. Many of those outside the analytics community wouldn’t know about all the websites and databases which are enormously valuable:
Corsica: Corsica is the main database used by stats people to get their data. They have tons of advanced stats for skaters and goalies, as well of useful charts, a similarity calculator and many, many other features. It is also very user friendly and comes with a useful glossary to explain stats.
Own the Puck: This a blog started by Domenic Galamini. This site is enormously valuable because it features HERO charts. Anyone who is active on social media has probably seen many HERO charts. They are useful tools to get a very good idea of how good a player is.
Puckalytics: This is another useful stat database similar to Corsica. The main feature that Puckalytics has that Corsica doesn’t is the Super WOWY which is a useful tool for looking at players with and without specific teammates.
Dispelling Voodoo: This is a blog on goalie stats by Ian Fleming and is a great resource. On this site you can find SAVE charts, which are similar to HERO charts, but for goalies. He also has charts on age progression which are interesting to browse. One of the top sites for goalie nerds.
Hockey Graphs: in terms of blogs about hockey analytics, there is no collection of smarter hockey minds than the writers at Hockey Graphs. This isn’t a statistical resource but is a phenomenal site to browse for those looking to find out more about hockey analytics.
Now if you have made it this far into this rather dry article, you’re probably at least somewhat intrigued by the concept of hockey analytics. I am not an expert on the topic, but if anyone has any questions, I can guarantee that I will answer to the best of my abilities or seek out someone more knowledgeable and get back to you. If anyone has any feedback, comments or questions, do not hesitate to send me a tweet (@CorsiGuy) or send me an email (firstname.lastname@example.org)