2022 JFreshHockey NHL Player Card Explainer

A step-by-step breakdown of how to understand all the stats on the JFreshHockey player cards.

Today is an exciting day for me: with the half-season threshold (finally) crossed and the sample sizes large enough to be meaningful, the time has finally come for me to release the 2021-22 Player Cards to Patreon subscribers. As of now, there are 713 player cards - 431 forwards, 209 defencemen, and 73 goaltenders - for this season, although that number grows every week.

The “player cards” are the core hockey visualization I’ve offered to subscribers of my Patreon for the past year and a half or so, and they’ve gone through a few changes over that time. Their objective is to present a diverse set of relevant statistics all in one place, giving you a comprehensive overview of a player’s bio, context, and performance in just one image. You might have seen them on social media either by following my Twitter account (@JFreshHockey) or seeing somebody else share them, but maybe you don’t know what they actually mean or how to read them. I think it’s worth another look at what exactly these stats mean, how they’re comminicated, and how to interpret them.

On a fundamental level, blue = good, red = bad. But if you’d like to go a bit deeper than that, hopefully this article will illuminate what exactly is being displayed here.

What is WAR?

At the basic level, most of what you’re seeing on these cards is based on the metric Wins Above Replacement, which originates from baseball but has been adapted to hockey by several analysts including Manny Elk, Dawson Sprigings, the Younggren twins, and Patrick Bacon, whose model is the one I use. For a thorough deep dive on Patrick’s model, refer to his overviews on Medium. But I’ll give you the gist here.

Wins Above Replacement, or WAR, is a statistical model designed to estimate the individual impact a player has had on his team in a number of ways, including by creating scoring chances for himself and his linemates at even strength and the powerplay, by finishing those chances, by preventing his opponent from getting scoring chances at even strength and the penalty kill, and by drawing more penalties than he takes. The model does this by using a technique called RAPM to isolate for a player’s teammates, opponents, deployment, zone starts, and other external factors that might affect their on-ice results. The output is an estimate of how many more games a player’s team won than they would have if they had instead played a “replacement level” player in those minutes. A replacement-level player is a fringe 13th forward or 7th defenceman, the kind of depth player that teams could easily acquire for almost nothing. For example, if the Oilers finish with 94 points and Connor McDavid was worth 8 WAR, the model is suggesting that if they had cut McDavid and played, say, Matt Nieto instead, they would have finished with 78 points instead.

WAR is by no means perfect, and as with any hockey model, the number that comes out should not be taken as a definitive measure of a player’s quality. While the metric has outperformed more traditional player evaluation tools like points in terms of describing and predicting team success, one of the great things about hockey is that there is still plenty of room for grey area and debate. If a fourth liner puts up great WAR numbers it doesn’t guarantee that he’s secretly one of the league’s best players and a coach is stupid for not playing him on line 1, but it could indicate a potential breakout - Andrew Mangiapane and Devon Toews were high-end WAR players long before they were household names.

There’s another important thing to consider here: WAR is a measure of performance, not ability. While we can use it to project a player’s future performance, talent isn’t something that can be easily quantified. The fluidity of hockey also means that a player might struggle or thrive in a certain environment or system and not in another. They also train in the offseason to address certain shortcomings, or their coach might adapt to them - and of course age and development can have both positive and negative impacts on how a player’s performance changes, sometimes suddenly. Past data suggests that switching teams does not significantly shift the model’s predictiveness, but that doesn’t mean external unmeasurable factors can’t influence a player’s results. In hockey, you are always wrong about most of your predictions no matter whether you use your gut, the eye test, or analytics. That’s just the way things go - even the best standings predictions in the world are off by an average of 7 points per team. The goal, as Patrick puts it, is “to be wrong less frequently.” The player cards will be most useful when combined with a thoughtful interpretive analysis (but that doesn’t mean they should just be dismissed out of hand if they don’t agree with one’s prior opinion).

How Do I Read the Cards?

Hockey fans have a good idea of how many goals or points is “good” or “elite” - 30 goals is a great season, 50 is elite, etc. The same isn’t true for WAR for pretty much anybody, even the most stats-savvy. To deal with this, instead of putting the raw numbers on the cards, I use percentiles. If a forward is 98% in EV Offence, for example, that means they rank higher than 98% of their peers in that stat - they’re in the top 2%. It’s why blue = good, red = bad.

Presenting the data like that has trade-offs. On one hand, it makes things a lot easier to understand for the vast majority of hockey fans than using raw data or z-scores, which is my primary goal. On the other, it does flatten things out at the extremes and make elite players look much closer than they are to the pack. For example, the top-rated forward in the league is projected to provide 8.2 wins per 82 games. The 10th-best is projected to provide 4.1, and the 324th-best is projected to provide 0. So the gap between #1 and #10 is the same as the gap between #10 and #324, but by percentiles it would look like 100%, 98%, 20% respectively. So it’s very important to keep in mind that at either end of the spectrum things get more extreme.

Okay, so let’s just go bit by bit here.

Forward Cards

Let’s walk through all the information on the forward player cards.

These provide context for a player as well as biographical information:

• Position: The position the player plays according to CapFriendly (the only site online that has accurate data on this)

• Age: Age as of the beginning of the 2021-22 season.

• Time-On-Ice: Time on ice per game translated into their role

• Cap: The cap hit and years remaining on their contract, per CapFriendly

• Competition: The percentile ranking of Quality of Competition a player faces on average, based on time-on-ice. A higher ranking means he is deployed against higher lines, and the inverse. This is already accounted for by WAR, so it should just be taken as insight into how a player is deployed by his coaches and matched against by opposing coaches.

• Teammates: The percentile ranking of Quality of Teammates a player plays with on average, based on time-on-ice. A higher ranking means he frequently plays with his team’s top TOI players, while a lower one means he’s probably in a depth or fourth line role. Similarly is already accounted for by WAR.

The following categories are percentile rankings among all forwards with 200 or more 5v5 minutes played in the 2021-22 season. All WAR stats are based on weighted averages of the past three seasons of data, with weights chosen to maximize predictiveness. 82-game paces are also used for everything so that injured players are not penalized.

• EV Offence: Projected Even Strength Offence WAR, meaning the wins a player provides to his team through his isolated impact on scoring chances (expected goals) for when he’s on the ice at even strength.

• EV Defence: Projected Even Strength Defence WAR, meaning the wins a player provides to his team through his isolated impact on scoring chances (expected goals) against when he’s on the ice at even strength.

• PP: Same as EV Offence but on the powerplay. If the player plays fewer than one minute per game on the powerplay it is listed as NA.

• PK: Same as EV Defence but on the penalty kill. If the player plays fewer than one minute per game on the penalty kill it is listed as NA.

• Finishing: The projected Wins Above Replacement a player provides by finishing his scoring chances, measured by goals scored above expected. This is a very important stat for forwards because it’s the most direct ways they create goals and therefore wins for their team compared to a replacement-level player.

• Penalties: The projected wins a player provides through his penalty differential on non-offsetting penalties. If you draw more penalties than you take, you’re helping your team and rank more highly.

• Proj. WAR %: The sum of all of the above numbers, expressed as a percentile.

The graphs on the right side of the player card show timelines of a few of these stats over the past three seasons, so you can get a sense of how things have changed in recent years. The top one shows percentile rankings of total WAR, the bottom one compared percentile rankings for offence, defence, and finishing.

Finally there are two more stats included:

• G/60: A player’s percentile ranking in projected goals per 60 minutes at 5v5, based on a weighted average of the past three seasons.

• A/60: The same, but with 5v5 primary assists per 60.

Defence Cards

The information displayed on the defence cards is the same as on the forward cards but with one major exception. As mentioned above, WAR values finishing very highly. But for defencemen, with very few exceptions there’s almost no year-to-year repeatability to that stat because of the nature of shots blueliners tend to take; think of Morgan Rielly’s out-of-nowhere 20 goal season in 2019 or Jakob Chychrun and Jeff Petry last year. While finishing does appear on the defencemen cards, it is removed from the final projected WAR calculation.

Goalie Cards

Goalie cards present stats the same way that the player cards do, but the stats are very different. Goalie WAR is a lot more straightforward than player WAR because the job they do is much simpler: they create wins for their team by keeping the puck out of the net better than a 3rd-stringer would. To make sure that we’re not crediting goalies for the team defence in front of them, goals saved above expected (GSAx) is the main input here. Whereas save percentage treats every shot as equally dangerous, GSAx weighs every shot based on the likelihood of it going in on average based on past seasons’ data. The sum of all the expected goals a goalie has faced compared to the number of goals he actually allowed gives us a number that more fairly reflects his actual performance.

Here’s a run-through of the info on these cards.

These provide context for a player as well as biographical information:

• Age: Age as of the beginning of the 2021-22 season.

• GP%: The % of his team’s games a goalie has started in the 2021-22 season

• Role: Role on the team based on GP% - Starter, 1A, 1B, or Backup

• Cap: The cap hit and years remaining on their contract, per CapFriendly

The following categories are percentile rankings among all goalies with 20 or more games played in the 2020-21 and 2021-22 seasons combined. All WAR stats are based on weighted averages of the past three seasons of data, with weights chosen to maximize predictiveness. Stats are on a per-game basis, regressed based on total GP in the past three seasons.

• EV WAR: Projected Even Strength WAR, meaning the wins a goalie provides by preventing goals against at even strength.

• PK WAR: Projected Penalty Kill WAR, meaning the wins a goalie provides by preventing goals against on the PK.

• High Danger: Goals saved above expected on high-danger chances

• Med Danger: Goals saved above expected on medium-danger chances

• Low Danger: Goals saved above expected on low-danger chances

• Rebound Control: Rebounds conceded below expected based on the quality of shots faced

• Consistency: The variation between WAR per 60 minutes from season-to-season - basically how predictable a goalie has been year-in year-out

To look at game-by-game performance, we use a variation on Robert Vollman’s “quality starts” stat - while that measure used league-average save percentage, we use goals saved above expected. Remember that these are percentile rankings where a higher % is better, not the actual % of quality/excellent/bad starts.

• Quality Starts: Frequency with which a goalie records a goals saved above expected above 0 in games they start

• Excellent Starts: Frequency with which a goalie saves 2 or more goals above expected in a game they start

• Bad Starts: Frequency with which a goalie allows 2 or more goals above expected in games they start

The graphs on the right side of the goalie card also show timelines of WAR per 60 minutes as well as a comparison of expected save percentage and actual save percentage.

Conclusion

Hopefully you now have all the info you need on how to fully understand and interpret these player cards! If you want to see these cards for all 713 current players and 8,035 player cards from 2007-2020, you can subscribe for $5 a month or more to my Patreon. It’s the easiest way to directly support my work, and you get access to an expansive set of visualizations that also includes team cards and more. All cards are updated daily, and will be for the remainder of the season.