Multi-Year Player Cards
Welcome to the main tool for the LB-Hockey project! The multi-year cards are the foundation for all the site’s dashboards as they display most of the data used throughout in detail. They look to measure a combination of a player’s impact and perceived skill through a combination of play-driving metrics & microstats. This allows us to quantify a player’s usefulness, along with its stylistic profile by displaying the various tools in its arsenal.
The skater or goalie view can be selected by switching between the two tabs at the bottom of the sheet. When inspecting the viz, one can change players by selecting the drop-down above the headshot and scrolling for or typing out the wanted player. A second customization option is available by choosing to display either their raw SPAR or the associated percentile (SPAR%). If the non-percentile SPAR option is chosen, the category scores will also change to showing how much SPAR each one contributed to the skater’s overall impact. Either way, the donut chart surrounding the number will convey the score’s percentile. The rest of this page will act as a glossary for these cards. More information on accessing & using it is available on the Registration page.
The Data
First, let’s establish some basic properties of the data. These multi-year cards cover a three-season span, with a heavier weight being accorded to the more recent seasons. As indicated, the raw data is pulled from a few sources, namely NHL.com, MoneyPuck (https://moneypuck.com), and Corey Sznajder’s All-Three-Zones Project (https://www.allthreezones.com). I highly recommend supporting these last two creators as they all do an incredible amount of work that is indispensable to the products on this site, and even more importantly, to the hockey analytics community as a whole.
Aside from a few exceptions (finishing is taken at all strengths for example), the skater metrics are calculated from 5-on-5 data and as rates to account for ice time disparity. They are weighed and summed to calculate categorical scores, and combined with power play, penalty kill, & clutch scores for the overall model SPAR (Standings Points Above Replacement). There is also an “Impact on Teammates/Opposition” component that is included in this calculation that rewards players who positively impact teammates’ expected goal shares and negatively so for opposing players using a “With or Without You” analysis.
The aforementioned weights applied to each category were empirically determined using their correlation to a team’s success and place within a skater’s toolkit. Training was completed against a team’s SPAR contributions coming from its forwards and defencemen, with the data covering 2021 to 2025. The weights were based on three factors: impact on team SPAR, year-over-year repeatability within the skaters, and multicollinearity with the other skills. It’s also worth noting that these final values differ by position, so entry chance defence is more important to Ds than Fs, which matches the intuition. All of these combine to get weights within and across categories which best estimate a player’s projected skill and contribution to their team.
Once a player’s skills weighted average is calculated, that is then converted into points probability added (above replacement). In other words, that’s the level to which a team’s PTS% would be expected to improve in a given game from having a player with those capacities and talent level added to their roster. We can easily expand that to a full season to then obtain SPAR. You can even view the SPAR numbers broken down by category in these cards, simply through selecting the aforementioned “SPAR” drop-down and shown in the picture above.
Each individual skill stat along with the five categorical scores are displayed in percentiles which, although provides an easy way to interpret stats relative to the rest of the league, also has its flaws. Therefore, it is important to remember how stark the difference can be at the top and bottom of a percentile scale. For example, a 100th percentile skater might contribute 3.5 more SPAR than one from the 98th percentile, while the difference between the 70th & 30th percentiles is only 2 SPAR.
The Layout
We will now tackle the different components displayed on the cards. What likely captures your attention at first are the badges. These have the goal of quickly informing you on the selection’s playing style. The titles should be fairly self-explanatory in terms of what they are meant to convey. Each badge has four levels: platinum, gold, silver, & bronze, with the following thresholds:
– Platinum: best forward and defenceman in the dataset
– Gold: next best 7 forwards & 4 defencemen
– Silver: next best 14 forwards & 8 defencemen
– Bronze: next best 21 forwards & 12 defencemen
These thresholds may be slightly smaller for a select few badges depending on their nature (ex: the Stylistic Unicorn badge is distributed to fewer skaters).
A quick explanation of what each badge is attempting to capture can be found at the badge glossary page. While a link to all the assets used in creating their visuals is hosted here.
Moving to the right column, we have the similar styles section. Here we intend to pair the selected skater with its three most similar (and single least similar) counterparts stylistically. This is done by comparing exclusively their metrics distribution, no effort is made to match talent level. The result is linking skaters that have a similar style of play but might not be succeeding with it at the same calibre. While the goalie cards still display their most similar equivalent, most of this column is dedicated to ideal systems. In a similar fashion to the skaters, the distribution of player stats is examined, but this time is compared to the volume of shot types associated with those skills. For example, goalies who score high in “Scramble Ability” but low in “Agility” will be matched with teams that may give up more scramble shots/plays but limit cross-crease one-timers.
Speaking of ideal complements, that is what’s being displayed on the bottom row of the skater cards. It works very similarly to the previously explained “Ideal Systems” algorithm. Using the eight categories demonstrated in the Comparison Atlas, linemates are found by looking at others who can mask weaknesses and support their strengths. Intuitively, a puck-carrying playmaker with little checking or defensive impact would thus end up paired with a responsible two-way power forward that can reliably finish their chances. Contrarily to the “similar styles” section, players with somewhat close ice time and overall talent level are prioritized, so as to mirror likely line combinations.
As for the goalies, the bottom strip displays their game quality spread. This aims to help visualize their consistency by seeing how often they have awful to incredible games. Since the model uses a GSAR (Goals Saved Above Replacement) approach, the quality of games is classified as follows:
– Awful Games: under -2 GSAR
– Poor Games: between -2 & -0.5 GSAR
– Fine Games: between -0.5 & 0.1 GSAR
– Good Games: between 0.1 & 1 GSAR
– Great Games: between 1 & 2 GSAR
– Incredible Games: over 2 GSAR
The Skaters
Now that the rest of the cards have been explained, I’ll give point-form explanations for each of the skater’s categories and skill metrics contained within.
Deployment: How a player is used & deployed by the coaching staff
– 5on5 TOI: Volume of 5-on-5 ice time given to the player
– Favourable Zone Starts: Share of shifts starting favourably (in o-zone) vs unfavourably (in d-zone)
– Power Play Time: Volume of power play time given to the player
– Penalty Kill Time: Volume of penalty kill time given to the player
– Quality of Teammates: Average strength of teammates when on the ice (weighed by ice time shared)
– Quality of Competition: Average strength of opposition when on the ice (weighed by ice time shared)
Zone Offence: How effective a player is in the offensive zone
– Chance Generation: Ability to generate chances for their team when the player is on the ice (adjusted for zone deployment and teammate/opposition strength)
– Sustained O-Zone Pressure: Ability to instil & extend offensive pressure, i.e. pouring on shots and holding control in the o-zone, when the player is on the ice (adjusted for zone deployment and teammate/opposition strength)
– Playmaking: Ability to create chances for teammates with passing plays (weighed by how dangerous these chances are)
– Individual Creation: Ability to create chances for themself off the forecheck and cycle (off in-zone plays)
– Finishing: Ability to outperform expected goal outputs on shot attempts
Zone Defence: How effective a player is in the defensive zone
– Chance Suppression: Ability to suppress the opposing team’s chances when the player is on the ice (adjusted for zone deployment and teammate/opposition strength)
– Mitigated D-Zone Pressure: Ability to limit opposing teams’ offensive zone pressure and prevent long possessions against, when the player is on the ice (adjusted for zone deployment and teammate/opposition strength)
– Retrievals: Ability to retrieve pucks in the defensive zone
– Puck Management: Ability to avoid turnovers with the puck in the defensive zone
– Exits: Ability to break the puck out of the defensive zone (weighed by how likely the attempt is to turn into an offensive zone possession)
Transition: How effective a player is in transition plays (mostly starting in the neutral zone)
– Entry Volume: Ability to frequently generate controlled entries through carrying or passing the puck past the blue line
– Entry Efficiency: Ability to enter the zone with possession on a good portion of entry attempts
– Entry Offence: Ability to create chances through passing or shooting on transition plays (weighed by how dangerous these chances are)
– Entry Disruption: Ability to stop the opposition from advancing zones in transition
– Entry Chance Defence: Ability to prevent the opposition from getting chances off transition plays
Checking: How a player affects their opposition (mostly through contact), and makes life harder for them
– Forecheck Pressure: Ability to apply pressure on opposing breakouts and stop exits in their track
– Forecheck Battling: Ability to recover dump-ins and steal pucks on forechecks
– Stick-Checking: Ability to strip opposing players of the puck
– Net-Front Play: Ability to outplay opposing players in front of the net to either generate chances there (as a forward) or prevent them (as a defender)
– Play Through Contact: Ability to engage in high-contact areas and successfully make plays when pressured
Teamplay: How a player affects their teammates, and makes life easier for them
– Teammate Utilization: Ability to make use of their teammates often and all over the ice when they have the puck through passing
– Importance to Teammate Offence: Ability to be an integral part of their team’s offensive network, in other words, having their teammates’ offence flow through the player (measured by how much it would negatively influence their teammates if the player was removed)
– Off-Puck Offensive Support: Ability to often be open off-puck as an option for teammates in the offensive zone
– Penalty Differential: Ability to maintain a favourable ratio of drawn and taken penalties
– Consistency: Ability to put together reliable performances day in and day out
The Goalies
When it comes to the goalies, measuring individual skills ended up being a far simpler approach since it essentially comes down to filtering for a certain type of shot on goal and measuring performance on that selection. This performance is measured by taking the goalie’s average goals saved above expected per xG. Moving to the five “highlight stats” in the centre bottom of the cards, it doesn’t make as much sense to give categorical scores so they have been replaced. The best & worst shots are evidently the types of shots against which they perform best & worst.
Next up is the Stolen Games square. In these multi-year cards, this indicates how many times over the last three years this goalie has saved more goals than their team’s win differential. In other words, a goalie will be credited with a stolen game if their GSAx is higher than how many goals their team won by (excluding empty netters). This concept can be easily flipped to get Choked Games too.
You might notice that these are different from the amounts provided by other models. This is because I have decided to exclude expected goals from missed shots since the goalie cannot be credited for having stopped them. As a result however, the expected goal calculation takes this into account and has its probability calculated given that the shot is on net.
This line of thinking is also applied to the final stat here: GSAR/GP. It is essentially the main overall performance metric I use for goalies (and which is then converted to SPAR using pythagorean expectations in win probability). The objective of this GSAR/GP statistic is to, based on their observed performance, conceptualize how many goals a goalie would be expected to save in an average NHL game, relative to a replacement-level goalie. By “average NHL game”, we imply NHL average shot counts and quality for the given season. Lastly, I should note that there is a year-over-year consistency factor and games played regression threshold that are applied to the GSAR/GP to SPAR conversion which explains why the rankings of these two metrics aren’t quite the same.
Mini-Cards
Mini-cards are also available in a separate tab, which are condensed versions which are better suited for social media thanks to their square dimensions and improved digestibility! They utilize the same SPAR model data, just trimmed down and formatted more concisely. Single-season views are available as well and are held in the same sheet as the full-sized single-season cards. One extra feature that these single-season ones have, is the ability to view the progression for each skill category by double-clicking on the timeline plot’s y-axis.
Closing Thoughts
I know it’s a lot but that’s what these cards are meant for: a more in-depth look at what a player brings to the table. They are far from perfect and continue to evolve. I first created these in late August 2022 and wanted to give them enough time so I could evaluate which metrics properly fit and conveyed what I wanted them to. Model improvements have come over time as a result but there are still holes as is the case everywhere. For example, forwards who are especially strong finishers but have a poor defensive impact and middling passing & transition results are undervalued in these cards, but at least still have their stylistic profile properly captured I feel. I continue to tune these cards as I have been to further optimize the end product and hopefully provide you all the best hockey analytics experience I can. Thank you so much for reading this glossary and I hope you enjoy all the site’s tools.
Cheers!
– Louis