[text_output]Hockey analytics continue to be a polarizing topic among hockey fans and media. While there are a multitude of reasons as to why some fans aren’t big proponents of the advanced statistical measures available today, one particular issue continues to rise to the top: hockey analytics can be very difficult to understand and contextualize. In fact, Washington Post sportswriter and self-proclaimed “stats geek” Neil Greenberg recently (July 23, 2018) sparked a debate on Twitter with this tweet:[/text_output][image type=”thumbnail” float=”none” src=”2967″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][text_output]While I staunchly disagree with the notion that “we’ve had nothing but Corsi-based stats for almost a decade,” he does bring up a very valid point at the end of the tweet regarding the lack of context that accompanies the various graphics and data visualizations that get bandied about social media and blogs. While it isn’t feasible for fans or analysts to attempt to fully explain every graphic every time they tweet one out, particularly for more data-rich visualizations such as Bill Comeau’s SKATR player comparison tool, there is absolutely a need for further context and explanation for many of the excellent player analysis visualization tools that are available to the public.

Consider this article my attempt to try to bridge this gap a bit, and provide that additional context and explanation that individuals like Neil Greenberg are looking for. In this piece, I will provide explanations for how to access, use and interpret a variety of the more popular hockey data visualization tools that are publicly available. If you still have questions about any of the charts I highlight, or if I left one out that you were hoping to learn more about, please don’t hesitate to reach out to me on Twitter, and I will do my best to answer any questions you have. If there is enough demand for it, perhaps I’ll write a part-two that provides explanations of a handful of other great hockey charts and data visualization tools out there.[/text_output][custom_headline type=”center” level=”h5″ looks_like=”h5″ accent=”true” id=”” class=”” style=””]SKATR Comparison Tool[/custom_headline][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]What is it?[/custom_headline][text_output]The SKATR player comparison charts, courtesy of Bill Comeau, are comfortably among the most popular among hockey fans on Twitter, and I honestly can’t recall the last time a transaction involving an NHL player occurred that wasn’t shortly thereafter accompanied by a tweet including that player’s SKATR chart. This data-rich chart is a personal favorite of my own, as it allows you to obtain a relatively intuitive, side-by-side comparison of two skaters at the same position. However, it goes beyond that, and allows you to view a single player at a time (presents the same data as the comparison feature), search for a player who reaches specific statistical benchmarks and craft full lineups and view how each included player performs across nine key statistics. For the purposes of this article, I will focus on the SKATR comparison tool feature, but again, please don’t hesitate to reach out to me if you have questions about the other features provided by Bill.[/text_output][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]How do I use it?[/custom_headline][text_output]The SKATR comparison tool generates a side-by-side comparison of any two selected skaters that play the same position; in other words, you cannot compare a forward to a defenseman. The top of the screen features dropdown menus that allow you to select the position pool (forward or defenseman), the seasons by which to compare players—current options are 2017-2018, 2016-2017 and 2016-2018—and the players to compare. The chart below dynamically updates, meaning that as you change the player or season selections at the top, the chart updates automatically, without having to press a submit button.[/text_output][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]How do I interpret the chart?[/custom_headline][image type=”thumbnail” float=”none” src=”2969″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][text_output]First, there are a couple of items you should be aware of before we get into the actual data contained in the chart. All of the data is also pulled from Corsica, which is important to note because some of the data points, such as expected goals and relative to teammate Corsi, have multiple models that exist, Corsica being just one of them. All of the data provided is 5v5 only, score and venue adjusted and presented as per 60-minute production. The reasons for the adjustments is to account for home ice advantage (venue adjustment) and to account for the score state during which a stat was recorded (score adjustment) because of how vastly different many NHL teams play when they are leading, trailing or tied. The tool uses per 60-minute production in order to help account for variations in ice time between two players, as a forward who logs 18 minutes of 5v5 time a game will be at a natural advantage in the counting stats like goals and points compared to a forward who only logs 10 minutes a game. Also, the player pool included in the tool consists only of skaters that amassed at least 100 minutes of 5v5 ice time during the selected timeframe.

When reading the chart, it is critical to know that the bars and numbers represent the percentiles, and not actual production. The length of the bar for each stat, the color of the bar and the number presented within the bar all represent the percentile the player falls into for that stat, relative to the other players at his position during the selected timeframe. Each white line on the chart represents a 20-percentile increment, and better ratings are represented by long bars, dark blue colors and high percentile numbers, while poor performance is shown by a short bar with a low percentile number and dark red color.

So, in the screenshot below which compares Chris Kreider (left) to Patrick Kane (right) across the prior two seasons, Kreider having a long, dark blue Points bar with the number 88 in it means that during the chosen timeframe, Kreider performed in the 88th percentile among NHL forwards at 5v5 play. It DOES NOT mean that Kreider logged 88 points during this timeframe. The chart does however list the player’s name and seasons being used for comparison just above the chart, and on the same line it also lists the player’s team, age, handedness, minutes played, goals, points and Corsi for percentage during 5v5 play.

Now that we know how to use and read the chart, let’s discuss the data actually contained in it. For the sake of brevity, I’m going to quickly discuss each statistic and what it measures, but if you would like a more thorough explanation, I encourage you to check out my Hockey Lexicon, which contains full definitions for all of the stats included in the chart as well as usage examples for many of them. Also, you will notice that many of the stats are underlined in this article; reason for that is that this website has a glossary feature, and when a term that appears in the glossary is typed for the first time in an article, it will be underlined, allowing you to tap (mobile devices) or hover-over (computers) the term to see a tool tip window containing a definition of the term.[/text_output][text_output]The first group of stats within the Individual section contains nine statistics that the individual skater accumulated himself (reminder that all stats are 5v5 only and per-60 minutes of playing time):

  • Game Score – One of my personal favorite statistics (and one I wrote an entire article about), Game Score is a catch-all statistic created by Dom Luszczyszynof The Athletic that quantifies the total value of a player’s productivity from a single game. It incorporates he following stats in an attempt to quantify the overall performance of a player: goals (G), primary assists (A1), secondary assists (A2), shots on goal (SOG), blocked shots (BLK), penalty differential (PD), faceoffs (FOW – FOL), Corsi differential (CF – CA) and goal differential (GF – GA). Obviously not all stats carry the same importance, so Dom assigned weights to each of the metrics to come up with the following formula for Game Score: Skater Game Score = (0.75 * G) + (0.7 * A1) + (0.55 * A2) + (0.075 * SOG) + (0.05 * BLK) + (0.15 * PD) – (0.15 * PT) + (0.01 * FOW) – (0.01 * FOL) + (0.05 * CF) – (0.05 * CA) + (0.15 * GF) – (0.15* GA)
  • Points – Standard box score points that all fans should be aware of that accounts for goals and assists.
  • Goals – Standard box score goals a player scored.
  • Prim Assist (primary assists) – The number of times that a player was the last person to touch the puck before another player scored the goal.
  • Sec Assist (secondary assists) – The number of times a player passed the puck to the player that gained the primary assist.
  • Indiv Shots (individual shot attempts) – Shot attempts taken by the player (including blocked shots and those that miss the net) that the goalie had to make a save on.
  • Indiv Exp Goals (individual expected goals for) – The amount of expected goals the player himself generated while he was on the ice. Expected goals is a statistic that considers both shot quantity and quality in order to provide a metric for how many goals a team (or player) should have scored, given the quality of scoring chances generated, if the opposing goalie played at a league-average level. Expected goals accomplishes this by weighting each unblocked shot attempt by a variety of shot attributes (e.g. shot type, distance, angle etc.), with heavier weightings applied to shot characteristics with a higher chance of leading to a goal.
  • Shoot % (shooting percentage) – The percentage of shots on net a player takes that are converted to goals.
  • Penalty +/- (penalty differential) – The amount of penalties a player draws minus the amount of penalties the player takes.

The second group of stats—On-Ice—features the following eight metrics that demonstrate how the player’s team performs while the player is on the ice:

  • Shot Share (CF%) (Corsi for percentage) – The percentage of all shot attempts (including blocked shots and shots that miss the net) taken by a team relative to the total amount of shot attempts taken by both teams while the player is on the ice.
  • Rel Teammate CF% (relative to teammate Corsi for percentage) – Relative to teammate metrics represent an improvement over standard (team) relative metrics, which simply measure how a team performs when the player is on the ice compared to when the player is off. I wrote an entire article on EvolvingWild’s teammate relative metrics, and while this tool uses Corsica’s data, the concept is the same. Relative to teammate statistics go a step beyond team relative metrics and attempts to further isolate a player’s performance by benchmarking his numbers against all of his individual teammates, weighted by the amount of ice time shared with each individual teammate, instead of against his entire team in aggregate. So, relative to teammate Corsi for percentage uses this type of calculation to try to isolate how impactful this particular player was on his team’s ability to generate shot attempts compared to the opponent, relative to all individual teammates.
  • Shots For (CF) (Corsi For) – The amount of shot attempts (including blocked shots and those that miss the net) a team takes while the player is on the ice.
  • Shots Against (CA) (Corsi Against) – The amount of shot attempts (including blocked shots and those that miss the net) the opposing team team takes while the player is on the ice. It should be noted that higher ratings here indicate fewer shots against, which is the positive outcome.
  • Exp Goal Share xGF% (expected goals for percentage) – The percentage of all expected goals generated by a team relative to the total amount of expected goals generated by both teams while the player is on the ice.
  • Rel Teammate xGF% (relative to teammate expected goals for percentage) – Uses the same relative to teammates calculation model as discussed in the relative to teammate Corsi for percentage definition. This stat measures the amount of expected goals a team generates when a player is on the ice compared to the total amount of expected goals generated by both teams, relative to all individual teammates in order to better isolate the player’s personal impact on the share of expected goals his team generates.
  • Exp Goals For (expected goals for) – The amount of expected goals a team generates while the player is on the ice.
  • Exp Goals Against (expected goals against) – The amount of expected goals the opposing team generates while the player is on the ice. It should be noted that higher ratings here indicate fewer expected goals against, which is the positive outcome.

The final group of stats—Context—consists of the following four metrics that help to provide important context that could impact all of the metrics in the Individual and On-Ice sections:

  • Share of Icetime (time on ice percentage) – The percentage of a team’s total 5v5 tome on ice played by the player.
  • Quality of Compet (quality of competition) – A measurement of the quality of competition the player faced. There are a few variations of quality of competition, but the ones in this chart use opponent time on ice as the barometer, because, in theory at least, the best players gain the most ice time. So, if the player in question is constantly on the ice when the opposing team has players that log heavy minutes each night, the player will have a higher quality of competition rating.
  • Qual of Teammates (quality of teammates) – Same concept as quality of competition, but this measures the quality of teammates a player is on the ice with. The higher a player’s quality of teammates rating, the more time they have spent on the ice with teammates who log heavy ice time.
  • Def Zone % Starts (defensive zone start percentage) – The percentage of all faceoff starts the player is on the ice for that are taken in their defensive zone. The higher the player’s defensive zone starts rating, the more often that player is being deployed in the defensive zone on faceoffs. It is important to note that, in general, the impact of zone starts are vastly overrated by many fans and analysts. While zone starts are important, it is important to understand that over 50% of player shifts begin on the fly (not after a stoppage of play) and there isn’t any tracking currently available to measure these on the fly shits (at least, none that have been popularized yet). So, because of this, zone start measurements only account for under half of a player’s total shifts, mitigating (but not completely erasing) their impact.

One last note about the statistics contained in the chart. If you’d like to view the actual data the player accumulated, as opposed to just their percentile ranking in their position, all you have to do is tap (mobile devices) or hover-over (computers) any stat, and a flyout window will appear that lists the 5v5 per-60 stats for every metric contained in the chart.[/text_output][image type=”thumbnail” float=”none” src=”2970″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]Usage Example[/custom_headline][text_output]Now that I’ve explained how to use the chart and I’ve defined all of the elements and statistics, I want to quickly mention how I personally use these charts to aide in my own player analysis. One thing that is critically important to understand is that some stats are more important than others. Sure, all of the stats have their value, Bill did not just randomly pick and choose stats to include in this viz. However, it would be a mistake to simply look at two players and assume that just because player A has more blue bars than player B, than that automatically means player A is the better player.

I don’t want this to turn into an argument over what stats are more important, as this isn’t the purpose of this article. However, for me personally, the first stats I look at in this chart while comparing players are Game Score, goals, primary assists, relative to teammate Corsi for percentage and relative to teammate expected goals for percentage. The reason for this is because I think Game Score is a fantastic catch-all metric that almost always passes the smell test in terms of accurately highlighting who the best players are, while the two relative to teammate metrics do a good job of isolating the player’s performance to demonstrate their true impact at generating shots and scoring chances when they are on the ice, relative to the opponent. Goals is obvious, as the whole purpose of hockey is to put the puck in the net, and I choose primary assists over points or secondary assists because they’ve been proven to be more repeatable that secondary assists, and by including both points and goals, you are double-dipping on the impact of goals while also including secondary assists. Now, to be clear, I look at all of the stats on the chart when I’m comparing players, but those five are the ones I personally weigh the heaviest in my analysis.

A good example of how you shouldn’t just simply look at this chart to see who has more blue bars to determine who is the better player can be seen when comparing Brayden Schenn (left) and Aleksander Barkov (right). Now, both players put up pretty impressive numbers last year and are on a similar tier across the board, so it’s really not that outlandish to look at this chart and pick Brayden Schenn. In fact, I showed this chart to a friend of mine who is unfamiliar with these charts which player he’d choose (I redacted the names), and he took the player on the left because, “he has less red bars, so my guess is he’s a more well-rounded player.”

However, for me, given the fact that I prioritize Game Score, goals, primary assists, and the two relative to teammate stats, I’ll take Barkov in this comparison. Barkov is at least 10 percentile points better than Schenn in relative to teammate xGF% and primary assists, has a slight advantage in Game Score, has equivalent relative to teammate CF%, and is slightly worse in goals. There are other reasons to like Barkov more, particularly his tremendous advantage in penalty differential, his much tougher deployment (yes, I said zone starts are overrated, but they still matter, particularly when there is this big of a gap) and his low shooting percentage could indicate that his goals percentile rank has much room to improve. I can’t blame my friend for choosing Schenn due to the fact that he has more blue bars, and that’s my whole point of this section. When you really analyze the graph, despite having a bit extra red in the mix, Barkov rates out as the slightly better player given the way I personally analyze these charts.[/text_output][image type=”thumbnail” float=”none” src=”2975″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”” style=””][custom_headline type=”center” level=”h5″ looks_like=”h5″ accent=”true” id=”” class=”” style=””]HERO Charts – Player Evaluation Tool[/custom_headline][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]What is it?[/custom_headline][text_output]Dominic Galamini Jr. recently updated his HERO Charts Player Evaluation Tool, which now depicts the scoring tier probabilities (left side of the chart) and shot impacts (right side of the chart) for each NHL player. The scoring tier probabilities are based off of the method he outlined in a recent Hockey-Graphs article on comparing scoring talent. If you want to know all of the intricate details of Dominic’s methodology, I highly suggest you read his piece. However, to quickly summarize the key points, it plugs players’ 5v5 primary points per-60 data into a complex statistical model that generates the probabilities of each player maintaining their prior level of play.

In his piece, Dominic lays out a comparison between the scoring rates (5v5 primary points per-60) of Nico Hischier (2.0) and Nikita Kucherov (1.8), and states that “Hischier has managed the higher scoring rate but that doesn’t convey much information without any notion of uncertainty – a significant issue considering that the chunks of evidence are of unequal sizes. It turns out that Kucherov has a sample that is about three times larger (3359 minutes vs. 1077 minutes) and so it is reasonable to expect that his observed scoring rate is likely more indicative of his true scoring talent. However, the degree to which we should feel more comfortable with the data being in Nikita’s favor and how that factors into our comparison of the two players is unclear.”

In other words, he’s saying that we can count on Kucherov’s scoring rate being indicative of his ability because he has a large sample size to back it up, and we don’t have a large enough sample size to know if Hischier’s scoring rate is truly sustainable, and we also don’t know how big of a factor the varying sample sizes should play. This is where the HERO Charts Player Evaluation Tool comes into play, as it generates its results using a complex model that Dominic created for the very purpose of accounting for these factors, and projecting forward the likelihood that a player will fall into a certain scoring tier (1st/2nd/3rd/4th line or depth player for forwards, 1st/2nd/3rd pair or depth player for defensemen).

As far as the shot impact portion of the chart is concerned, Dominic presents the per-60 minute offensive and defensive “usage adjusted shot impacts” for each player, relative to the league average at the position. The usage adjusted shot impacts are built off of similar principals as EvolvingWild’s RAPM data (which I wrote about here). Long story short, it’s built using rate adjusted plus minus (RAPM) metrics, which attempt to isolate a player’s performance and impact, independent of both his teammates and opponents. In other words, RAPM measures a player’s actual contribution to his team in the various statistic being discussed (in this case shot attemps) by stripping out factors outside of the player’s control, such as the strength of his teammates and strength of his opponents. These measurements are superior to team relative statistics (usually simply referred to as “relative” statistics) that you see thrown around Twitter and other channels, as standard relative metrics simply illustrate how a team does while a player is on the ice compared to off, which of course can be greatly influenced by the other players on the ice.[/text_output][image type=”thumbnail” float=”none” src=”2983″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]How do I use it?[/custom_headline][text_output]Using the chart is incredibly simple. The very top of the page allows you to toggle between forwards and defensemen; note that the box with the darker gray shade is the one that is currently selected. After you choose the position, you then use the dropdown menu beneath to select the player you wish to view. The dropdown contains all players included in the tool, and a search bar appears at the very top of the dropdown that allows you to search for a specific player. Once a player is selected, his data is automatically populated in the two side-by-side charts beneath. The Scoring Tier Probabilities section displays bar graphs and accompanying percentages that indicate the likelihood that the player will perform to the level of play of a particular line. The right side of the chart displays the offensive and defensive impact the player had on his team, as measured by usage adjusted shots per-60 minutes of play vs the average skater at his position. You can hover-over the offensive and defensive portions of the chart to see the numerical impact of each player.[/text_output][image type=”thumbnail” float=”none” alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=”” src=”2984″][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]How do I interpret the chart?[/custom_headline][text_output]It is important to note that this is a unique player evaluation tool in that the left side of the chart (Scoring Tier Probabilities) is forward-looking, while the right side (Usage Adjusted Shot Impacts) is backward-looking. The Scoring Tier Probabilities section is incorporating past scoring rate data into a statistical model to project how likely the player is to perform at certain levels in the future, while the Usage Adjusted Shot Impacts section displays the offensive and defensive impact the player had across the previous three seasons. Most of the popular evaluation tools out there display data on how players performed in the past; while some of the data displayed is more predictive in nature than others, they are not providing forward-looking projections. 

With that in mind, the interpretation of the chart should be as follows. The left side of the graph projects what line a player belongs on in terms of production given their prior scoring rates and the probability they will be able to sustain that production moving forward. The right side of the chart shows the impact the player has had in the past on both offense and defense, relative to league average at the position. While there certainly are more factors to offensive and defensive performance than shot metrics, even when they are adjusted for factors outside of the player’s control (e.g. usage and quality of teammates), these metrics do serve as a strong indicator for how impactful the player was in both aspects of the game.[/text_output][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]Usage Example[/custom_headline][text_output]Because of the way Dominic’s model works, I think a great usage case for his player evaluation tool is to look at younger players with relatively smaller sample sizes in order to try and understand how their production projects going forward. Will Butcher of the New Jersey Devils and Charlie McAvoy of the Boston Bruins both had fantastic rookie campaigns last year, landing each of them some Calder Trophy votes, with McAvoy finishing 5th in the voting and Butcher placing 9th. However, both young players only have one full season under their belts, which is surely enough to get excited about a player, but not enough to definitively state that they can continue this production going forward.

As you can see in the two charts below, both players are projected to be impactful NHL producers going forward, and each had decent offensive impacts and slightly negative defensive impacts vs league average. However, Dominic’s model strongly favors Charlie McAvoy in terms of scoring tier probabilities going forward, giving him a 64% chance to produce at the level of a top-pair defenseman going forward, with a 30% chance of producing at the level of a second-pair defenseman and a 6% chance of third-pairing production. While the model likes Butcher as well, it gives him the highest probability of producing at the rate of a second-pair guy (42%), compared to a 33% chance of producing at a top-pairing level, 23% chance of third-pair production and a tiny (2%) chance of being only a depth-level producer.[/text_output]

[image type=”thumbnail” float=”none” src=”2985″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””]
[image type=”thumbnail” float=”none” src=”2986″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””]
[custom_headline type=”center” level=”h5″ looks_like=”h5″ accent=”true” id=”” class=”” style=””]RAPM Player Comparison Charts[/custom_headline][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]What is it?[/custom_headline][text_output]While the topic of regularized adjusted plus minus (RAPM) data is fresh in our minds, lets transition to talking about how to use and interpret the Twins Josh and Luke (who share the Twitter account EvolvingWild, a must follow for any hockey fan) RAPM player comparison charts. Josh and Luke recently launched their new website (which I recently wrote about), and perhaps my favorite feature of the site is their RAPM player comparison charts. They currently have the even strength charts available, and will soon be launching a powerplay performance as well. The RAPM charts provide a side-by-side comparison of two selected players across five important RAPM metrics (per-60) at even strength: goals for, expected goals for, Corsi for, expected goals against and Corsi against. Using this tool, you can get a strong picture of the impact players had on their team in terms of generating goals, scoring chances and shot attempts, as well as their ability to suppress expected goals and shot attempts against.

In case you skipped over the last section on Dominic’s new HERO charts, or just have a poor memory like myself, here is a quick recap on what RAPM data is. RAPM attempts to isolate a player’s performance and impact, independent of both his teammates and opponents. In other words, RAPM measures a player’s actual contribution to his team in the various statistic being discussed (e.g. shot attempts, expected goals and goal differential) by stripping out factors outside of the player’s control, such as the strength of his teammates and strength of his opponents. I will again state that this sort of teammate relative data is superior to standard relative data (team relative), and I urge everyone to consider using teammate relative data instead of team relative moving forward.[/text_output][image type=”thumbnail” float=”none” src=”2991″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]How do I use it?[/custom_headline][text_output]After arriving on the Even Strength RAPM Charts page, you simply need to use the provided drop-down menus on the left to generate the desired player comparison, which appears to the right. The Player 1 drop-down displays in the left-side chart, while the Player 2 drop-down displays in the right-side. You can either select a player from the drop-downs or type in the player’s name you wish to view data for. The third drop-down allow you to choose the time frame for which to compare players, which are grouped into three-year increments. At this time, the only time frame choice is 2015-2018. One important note: unlike the previous two charts discussed, the results do not dynamically update as you make selections from the drop-downs, and you must click the blue Submit button in order to generate the results.[/text_output][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]How do I interpret the chart?[/custom_headline][text_output]The charts contain bar graphs for each metric that depict how players perform in each statistic, relative to the average player in terms of standard deviation. The standard deviations are charted on the y-axis, and the stat types are presented across the x-axis. The darker the blue and higher the bar reaches, the better the player is in the stat; the darker the red and lower the bar reaches, the worse the player is in the stat. The metrics provided in the charts are all per-60, even strength RAPM statistics and include:

  • Off_GF (goals for)– Goals scored by the player’s team while the player is on the ice.
  • Off_xG (expected goals for)– The expected goals for generated by the player’s team while the player is on the ice.
  • Off_CF (Corsi for) – The shot attempts generated by the player’s team while the player is on the ice.
  • Def_xG (expected goals against)– The expected goals generated against the player’s team while the player is on the ice.
  • Def_CF (Corsi against) – The shot attempts generated against the player’s team while the player is on the ice.

So, given all of this, the way you should interpret the chart is as follows. Having a long, dark blue bar means that the player has a strong positive impact on the team relative to the average player, independent of his teammates and opponents. Having a long, dark red bar means the opposite: a player has a strong negative impact on the team in the given statistic. So, if a player has all long, dark blue bars, you can interpret that to mean that even after accounting for his teammates and opposition, he has a strong positive impact on his team’s ability to score, generate scoring chance and generate shot attempts, as well as prevent the opposition from generating scoring chances and shot attempts.[/text_output][image type=”thumbnail” float=”none” src=”2988″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]Usage Example[/custom_headline][text_output]For this usage example, let’s take a look at two of my personal favorite non-Ranger players in the NHL: Johnny Gaudreau and Blake Wheeler. You can see from the chart that Gaudreau is approximately one and a half standard deviations better than the average player in terms of his impact on his team’s ability to score goals and generate expected goals, and had decently positive impacts with respect to generating and suppressing shot attempts. The fact that his expected goals bar far exceeds his Corsi for bar is an indication that Gaudreau is excellent at generating high danger scoring chances, and isn’t just padding his expected goals total by just getting anything and everything on net. Gaudreau does struggle a bit however in terms of limiting opponent scoring chances, as evidenced by his negative expected goals against bar.

Blake Wheeler however appears to have a more rounded game, with positive impacts across all five statistics. His biggest strengths lie in his ability to generate scoring chances and expected goals, as he is roughly one and a half standard deviations better than average at both He also is about one standard deviation better than average at suppressing opposing shot attempts, and has a positive impact of slightly under a standard deviation better than average on his team’s ability to score when he is on the ice. His worst stat, expected goals against, is impressively still about a half of a standard deviation better than average.

Long story short, what this comparison tells us is that over the past few years, Wheeler and Gaudreau are both excellent offensive contributors, but Wheeler has a decided advantage in terms of the impact he has on his team defensively.[/text_output][image type=”thumbnail” float=”none” src=”2990″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][custom_headline type=”center” level=”h5″ looks_like=”h5″ accent=”true” id=”” class=”” style=””]All Three Zone Player Comparison Tool[/custom_headline][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]What is it?[/custom_headline][text_output]CJ Turtoro created the All Three Zone Player Comparison Tool using passing and zone entry/exit data from Corey Sznajder. This tool does an excellent job of illustrating the impact that players have across four critical elements of the game: passing ability, zone entries, zone exits and zone entry defense. The tool provides a side-by-side comparison of two skaters across 10 total micro-stats that are bucketed into four categories measuring those elements of the game. These micro-stats provide fans and analysts with an objective view of the impact that players make in areas of the game that are more nuanced and difficult to track with traditional stats. It should be noted however that while very important, these stats are not meant to serve as an encapsulation of a player’s overall ability, so this tool should not be relied on to definitively state one skater is a better all-around player than another. However, this tool provides critical additional information that we can use when analyzing a player and comparing skaters.[/text_output][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]How do I use it?[/custom_headline][text_output]The top of the page features three drop-down menus for each player that allow you to choose which players and seasons to compare: Team, Season and Player. The Team drop-downs serve as a filter for the player drop-downs; after a team is selected, the player drop-down will only contain skaters that played for the selected team during the selected season. At the time of writing this, the Season drop-downs allow you to select from one of three options: 2017-2018, 2016-2017 and 2016-2018. The player drop-downs feature a search bar at the top that allows you to type in the name of the desired skater, or you can select the skater from the provided list. The charts below dynamically update as you make selections from the drop-down, so you do not need to click a submit button to generate the results.[/text_output][image type=”thumbnail” float=”none” src=”2992″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]How do I interpret the chart?[/custom_headline][text_output]All of the data provided in the tool is 5v5 from the 2016-2018 seasons and was collected by Corey Sznajder. It should be noted that this data is manually collected by Corey, and therefore it takes a tremendous amount of time and effort, so the data sample size is not a full 82 games per season. Each player section lists the games and minutes tracked for the player across the selected seasons, as well as the player name, team, position, seasons selected and a chart key.

Similar to the SKATR tool, it is critical to know that the bars and numbers presented within the chart represent percentiles, and not actual production. The length of the bar for each stat, the color of the bar and the number presented to the right of the bar all represent the percentile the player falls into for that stat, relative to the other players at his position during the selected timeframe. Better ratings are represented by long bars, dark blue colors and high percentile numbers, while poor performance is shown by short bars with a low percentile number and dark red color.

Now that you understand how to read the chart, let’s quickly discuss the stats involved. First up, we have the shot contributions section, which contains three stats that show how a player contributes, via passing and taking their own shots, to the shot totals a team accumulates while the player is on the ice:

  • Shot Contributions – The amount of shot attempts plus shot attempt assists (defined next) a player accumulates per-60 minutes of 5v5 ice time.
  • Shot Assists per-60 – The amount of times a player passed to the player who took a shot attempt, per-60 minutes of 5v5 ice time.
  • Shots per-60 – The amount of shot attempts (not just shots on goal) a player takes himself per-60 minutes of 5v5 ice time.

The second section, Entry, contains two stats the depict the skater’s ability to enter the offensive zone with possession of the puck:

  • Possession Entries per-60 – The amount of times a player successfully enters the offensive zone with control of the puck per-60 minutes of 5v5 ice time.
  • Possession Entry Percentage – The percentage of zone entries a player makes with possession, relative to all zone entry attempts, per-60 minutes of 5v5 ice time.

Next, the Exit section provides two stats illustrating a skater’s performance relative to exiting their own defensive zone with possession of the puck:

  • Possession Exits per-60 – The amount of times a player successfully exits his own defensive zone with possession of the puck per-60 minutes of 5v5 ice time.
  • Possession Exit Percentage – The percentage of times a player successfully exits his own defensive zone with possession of the puck, relative to all zone exit attempts, per-60 minutes of 5v5 ice time.

The final section, Entry Defense, only appears for defenseman and contains three stats the show how adept a defenseman is at preventing the opponent from entering the offensive zone:

  • Breakups per-60 – The amount of times a defenseman was able to break up the opposition’s attempt to enter the offensive zone per-60 minutes of 5v5 ice time.
  • Possession Entries Allowed per-60 – The amount of times an opponent was able to successfully enter their offensive zone while the defenseman in question was the primary target to defend the zone entry attempt, per -60 minutes of 5v5 ice time.
  • Possession Entry Percentage Allowed – The percentage time an opponent was able to successfully enter their offensive zone, relative to all zone entry attempts, while the defenseman in question was the primary target to defend the zone entry attempt, per-60 minutes of 5v5 ice time.

What does this all mean? If a player has all blue bars with high percentiles in the Shot Contributions stats, that means that they are an effective player at generating shot attempts for their team through either taking the shot themselves or by passing the puck to the teammate who then ripped a shot towards the net. This adds valuable additional data to simple shot attempt metrics, as it is a valuable skill to be able to consistently be the set-up man who is able to pass the puck to a player in position to take a shot attempt. The Entry and Exit sections measure significant components of neutral zone play: the abilities to enter the offensive zone with possession of the puck and the ability to exit their own defensive zone with possession of the puck. The Entry Defense section, which I repeat is only available for defenseman, shows a player’s ability to prevent the opposition from entering their offensive zone, a critical aspect to defense, as allowing the opposition to cleanly enter the offensive zone consistently will inherently lead to many more high-quality scoring chances.[/text_output][image type=”thumbnail” float=”none” src=”2993″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]Usage Example[/custom_headline][text_output]For our usage example, I’ll use two defensemen in order to be able to view the Entry Defense sections for each player. Here you see Kevin Shattenkirk on the left, and Adam Larsson on the right, with data for the prior two seasons being considered in the charts. In this comparison, you see that Shattenkirk is in the 60th percentile of NHL defenseman in total shot contributions per-60 minutes of 5v5 play, which consists of a 78th percentile ranking in shot assists but only a 37th percentile ranking in shot attempts. Comparatively, Adam Larsson is slightly worse than Shattenkirk in both shot assists and shot attempts, leading to Shattenkirk having a sizable advantage in total shot contributions, despite have a brutal season last year (Shattenkirk’s chart for his injury-plagued 2017-2018 season is pretty ugly).

You can also see the Shattenkirk is clearly the superior player in terms of gaining entry to the offensive zone with possession, a crucial ability for any “puck moving” defenseman. In terms of exiting the defensive zone, another critical trait of a “puck moving” defenseman, Shattenkirk also has the upper hand, but that largely is due to the fact that it appears he attempts to exit the zone with possession far more often than Larsson, as evidenced by Shattenkirk’s 93rd percentile rank in possession exits per-60 but only being in the 56th percentile in terms of success rate.

Most fans who have watched both players over the years shouldn’t be surprised by any of these stats, and they likely won’t be surprised by the fact that Larsson is clearly the more effective player in terms of guarding his own blue line. Shattenkirk has abysmal numbers across the board in terms of his ability to prevent the opposition from gaining entry with possession into the offensive zone, while Larsson ranks quite favorably, particularly in his abilities to break up opposition entry attempts and his success rate at preventing opposition entries. However, the fact that Larsson has a below-average possession entries allowed per-60 but a high success rate at preventing entries indicates that, for whatever reason, the opposition is frequently targeting him when trying to enter the zone.[/text_output][image type=”thumbnail” float=”none” src=”2994″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][custom_headline type=”center” level=”h5″ looks_like=”h5″ accent=”true” id=”” class=”” style=””]Player Traits and Performance Comparison Tool[/custom_headline][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]What is it?[/custom_headline][text_output]Last but not least for this initial foray into contextualizing some popular hockey analytics charts we have the Player Traits and Performance Comparison tool, courtesy of noted hockey statistician and writer for The Athletic Ryan Stimson. This tool combines statistical concepts we discussed earlier with EvolvingWild’s HERO Charts Player Evaluation Tool and the All Three Zone Player Comparison Tool and expands upon the All Three Zone Player Comparison tool to provide a more holistic view of a player’s performance and abilities. It provides a side-by-side comparison of the impact that two players have on their team’s defense and how much of a team’s offense flows through the players in question. Ryan wrote about this tool in a fantastic article he wrote for The Athletic in early July that, in addition to explaining this tool, discusses the importance of certain advanced stats and why passing data is important. I highly recommend checking out the article, but for the purposes of this piece, here is a very high-level summary.

Ryan notes that “how teams strategically and skillfully move the puck (or ball in other sports) is perhaps the single-most important thing they can do,” and he goes on to link to articles with supporting evidence of this across the NFL, soccer and the NBA. He links to a Twitter thread of his from June, where he links to articles that highlight the impact that skillful passing has on scoring, how certain passes consistently lead to higher-danger scoring chances, and the predictive value that passing data can have in terms of projecting player production. To make a long-story short, passing ability is critically important in hockey (and many other applicable sports), and he provides more than ample evidence to back up this claim.

In terms of the tool itself, it provides you with a variety of information, most notably the player’s impact on defense and shot attempt contributions, as well the expected primary points he should generate given his shooting and passing ability. The tool then also provides a variety of contextual metrics that help users understand the components that go into determining their impact on shot attempt contributions and expected primary points, such as how they rank across a variety of types of passes (e.g. passes that lead to one-timers), their individual shots and primary shot assists, and the types of shots they attempt.[/text_output][image type=”thumbnail” float=”none” src=”2995″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]How do I use it?[/custom_headline][text_output]If you’ve read this entire article up until this point, I understand that these sections are pretty much all identical, due to the fact that most of these tools are created in Tableau. However, I continue to include this for each tool assuming that many that read this are simply skipping down to the specific tool they want to learn about.

In order to generate a the player comparison, you have to use the drop-down menus at the top to select the players to use in the comparison and the seasons of data to include in the analysis. A Team dropdown menu also exists for each player which serves as a filter for the Player dropdown. Similar to other tools in this article, you can select a player from the dropdown list or type in his name to quickly locate the player. The options within the Season dropdown include: 2014-2015, 2015-2016, 2016-2017, 2017-2018 and 2014-2018. Similar to the other Tableu tools covered, the results dynamically update upon making new selections to the dropdown menus, so there is no need to click any sort of submit button.[/text_output][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]How do I interpret the chart?[/custom_headline][text_output]Because this tool is using manual tracking data from Corey Sznajder and Ryan Stimson’s Passing Project, the same disclaimer must be made that I stated in the All Three Zone Player Comparison Tool section: this data is manually collected and therefore it takes a tremendous amount of time and effort, so the data sample size is not a full 82 games per season. Make sure to note the Minutes Tracked portion for an understanding of the sample size, which is presented beneath each player’s name within a section that also lists the seasons used in the comparison, player position, data sources and a chart key. All of the data included in the tool is at the 5v5 game state only.

Interpreting the data presented in this tool is very similar to the All Three Zone Player Comparison tool. The bars and numbers presented within the chart represent percentiles and not actual production. The length of the bar for each stat, the color of the bar and the number presented to the right of the bar all represent the percentile the player falls into for that stat, relative to the other players at his position during the selected timeframe. Better ratings are represented by long bars, dark blue colors and high percentile numbers, while poor performance is shown by short bars with a low percentile number and dark red color.

In his article, Ryan Stimson notes that the top-three stats are by far the most critical in terms of comparing the abilities of two players, and he state in his article that, “those are really all you need to know for overall player value. I include everything else because my interest in getting into hockey analytics years ago was always about learning what players fit well together, specific way teams generate and suppress offense, and attempt to apply measurable and valuable metrics to tactics within games.” In other words, the first three statistics are the true measure of player value, while the remaining 11 stats are measures of how they accumulate that value, and information that can help us understand if two players’ game styles would mesh well together. For example, if one player has a proclivity for passes that lead to one-timers, and another player isn’t a high-volume passer but takes a lot of one-timer shots himself, they could be a good match.

Those all-important three stats, which are presented at the top of the chart within the Defense and Impact sections, include:

  • RelTCA60 – This stat, teammate relative Corsi against per-60, is a measurement of shot suppression. It measures the amount of shot attempts the opponent takes while the player is on the ice, per-60 minutes of 5v5 play, relative to the player’s teammates. In theory, if the opponent consistently takes less shots while player A is on the ice compared to when any other teammate of player A is on the ice, than he is defensing effectively and not allowing the opposition to get into a position to rip off a shot.
  • SCB% – To be completely candid, I don’t have the slightest clue what exactly the acronym SCB stands for, but in the grand scheme of things that doesn’t matter. What I can tell you with 100% certainty, and what does matter, is that this stat measures the percentage of shots attempts that a team takes while a player is on the ice which the player in question directly contributed to via taking the shot himself or setting up the shot with a pass.
  • xPrP60 – This stat, expected primary points per-60, is similar to the expected goal models that have become very popular within the hockey community, with one significant difference. An expected goal model measures the amount of goals that would have been scored, given league average goaltending, due to the quality of scoring chances generated by the team. This stat however, expands that, and measures the amount of expected primary points (primary assists + goals) a player should register, isolating for factors out of his control such as fantastic goaltending or a terrible shooter. This stat accomplishes this by weighting a player’s shot contributions by quality; shot contributions that have a higher likelihood of converting into a primary assist or a goal have a higher weighting than those with a lower likelihood of converting into a primary assist or a goal.

The remaining 11 stats are organized across three sections—Pass Qualities, Shot Creating and Shot Qualities—and include the following:

  • 1TSA60 – The number of times per-60 minutes of 5v5 play that a skater passes the puck to a teammate that rips a one-timer shot attempt.
  • BuildUp60 – This stat measures the “build-up” in play leading to a shot attempt. Specifically, it measures the secondary and tertiary shot assists per-60 minutes of 5v5 play that a skater registers.
  • DZSA60 – This stat might be by personal favorite among these 11 contextual metrics, as it measures the number of shot assists a player registers that lead to a high-danger scoring chance. Specifically, it measures the shot assists created from a pass from behind the net to a player in front of the net, or a shot assist created from a pass across the slot, which have been proven to lead to higher quality scoring chances.
  • ixA60 – Individual expected assists per-60 minutes of 5v5 play. Think of this as the same thing as expected goals, except it measures the amount of assists a player should accumulate given the quality of the pass and scoring chance.
  • Trans60 – This stat measures the amount of shots and shot assists a team takes per-60 minutes of 5v5 play due to a “transition” pass from the player in question. Specifically, it measures the shots and shot assists resulting from passes from the player from the neutral or defensive zone.
  • PSC60 – Primary shot contributions per-60 minutes of 5v5 ice time. Primary shot contributions include shots a player takes himself, and primary shot assists.
  • SA60 – The number of shot assists per-60 minutes of 5v5 play that a skater registers. Shot assists are when the skater passes the puck directly to the player that takes the shot attempt.
  • Shots60 – The amount of shots the skater takes per-60 minutes of 5v5 play.
  • 1T60 – The amount of one-timer shots a player takes per-60 minutes of play.
  • iDZ60 – The amount of times a player takes a high-danger shot that resulted in a pass from his teammate across the slot or from behind the net, per-60 minutes of 5v5 play.
  • ixG60 – The amount of expected goals the player himself generated per-60 minutes of 5v5 play. Expected goals is a statistic that considers both shot quantity and quality in order to provide a metric for how many goals a team (or player) should have scored, given the quality of scoring chances generated, if the opposing goalie played at a league-average level. Expected goals accomplishes this by weighting each unblocked shot attempt by a variety of shot attributes (e.g. shot type, distance, angle etc.), with heavier weightings applied to shot characteristics with a higher chance of leading to a goal.
[/text_output][custom_headline type=”left” level=”h6″ looks_like=”h6″ accent=”true” id=”” class=”” style=””]Usage Example[/custom_headline][text_output]Due to the fact that this tool contains stats that measure player value—RelTCA60, SCB% and xPrP60—and those that are descriptive of a skater’s playing style, this tool has two primary uses: comparing player value and evaluating whether two players could be good fits together on the ice. Because of this, we will use two players on the same team for this usage example: Ryan Ellis and Roman Josi.

Ryan Ellis is personally one of my favorite players in the NHL, and one that I also view as among the more underrated players. He’s a guy that many fans and acknowledge as a very good player, but he rarely gets considered in conversations regarding the best defensive players in the league; personally, I think Ellis is comfortably a top-20 defender in the NHL, and I’d probably put him closer to 10 than 20.

If you look at the three most important bars for determining player value in the chart below—the defense and impact stats—you will see a picture that likely doesn’t surprise you: Ryan Ellis is the far superior defensive player, but Josi has the advantage in terms of offensive impact. It should come as no surprise to anyone that Roman Josi is among the most impactful defenseman in the NHL in terms of offensive impact, as he is in the 95th and 94th percentile of defensemen in terms of shot contributions and expected primary points. Ellis is no slouch in these areas either, placing in the 82nd and 66th percentile in these metrics. In terms of defensive impact, while I was under no false illusions that Josi was a shutdown defender, I was surprised to see him as low as the 11th percentile in terms of teammate relative shot suppression. Ellis on the other hand is in the 75th percentile, which coupled with his offensive impact stats makes him a very well-rounded defenseman. In terms of overall value, it’s tough to use this tool to definitively state who the better player is; for me, it would likely come down to team fit and what I personally need on the blueline.

In terms of fit, when you analyze the trait stats in the pass qualities, shot creation and shot quality sections, you see these two are a great fit for one another. For example, Josi is in the 100th percentile of one-timer shot assists, but does not take many one-timer shots himself. Conversely, Ellis is in the 79th percentile among defenseman in terms of the amount of one-timer shots he takes. We can use this to say that due to these traits, they are a great fit because Josi looks to set up one-timers, while Ellis loves to take them. Another example can be seen in examining the high-danger shot and pass traits of each player. Ryan Ellis has a penchant for passing the puck across the slot or from behind the net to an open player who takes the shot, ranking in the 80th percentile among defenseman in this passing trait stat. Josi on the other hand has a proclivity for getting open in the slot in a position to receive this exact type of pass and then firing it on net, as he ranks in the 93rd percentile among defenseman in individual high danger shots.

Given each player’s strengths, weaknesses and traits highlighted in the tool, I would conclude that Ryan Ellis and Roman Josi are an excellent fit, and that combined with both players’ abilities is a big reason why they are among the most dangerous defensive pairings in the NHL.[/text_output][image type=”thumbnail” float=”none” src=”2996″ alt=”” href=”” title=”” info_content=”” lightbox_caption=”” id=”” class=”aligncenter” style=””]

Author: Drew Way

Diehard New York Rangers fan since 1988! Always has been fascinated by sports statistics, and is a big proponent of supplementing analytics with the eye test. Also a big Yankees, Giants and Knicks fan.