Patrick Kane and What His Offensive Success Means in a Broader Context

Patrick Kane has long been a polarizing figure in the NHL, but there's simply no denying his offensive ability. What conclusions can be drawn from how he finds success?

Three-time Stanley Cup winner. Hart, Art Ross, Lindsay, Calder, Conn Smythe Trophies. Named a top 100 player in NHL history.

It certainly appears to fit the resume of a Hall of Fame, superstar career. Very few players can pile up such an impressive assortment of accolades before even turning 30. Even today, there’s no denying Kane’s talent or game-breaking ability. Among the most exciting hockey players in the world, Kane catches the eye of any and all viewers whenever he steps on the ice. It is no surprise that he continues to be praised by proponents of the eye test. Certainly, at first glance, it is hard to pick at his flaws given his immense skill. Kane’s prolific scoring also contributes to his elite reputation among much of the hockey world. Over the past twelve seasons, Kane leads all players with 906 points. He is third in goals and second in assists over that same span. It is not as though Kane has slowed down in that respect either; Kane is fourth among all players over the past five seasons behind highly-regarded stars Connor McDavid, Nikita Kucherov, and Leon Draisaitl.

And yet, Patrick Kane continues to be heavily criticized by users of hockey analytics. I, myself, am no exception. From the moment I began to work with and value advanced statistics, I have been a major detractor of his. For all of his accolades and point totals, Kane remains weighed down by a dearth of elite, or even average ability at the other end of the rink. Kane’s defensive deficiencies are well-documented, so I won’t spend a lot of time explaining them. To give a brief overview, Kane ranks second-to-last among all players in Evolving Hockey’s Def and expected Def (xDef) since 2017, both of which attempt to quantify a player’s defensive impact using their results at even strength and on the penalty kill. He also consistently ranks at the bottom of the league in expected Goals Against per 60 (xGA/60), an RAPM statistic also from Evolving Hockey.

So to reiterate, there’s no denying Kane’s struggles on defense. He is a liability in the Blackhawks’ zone and among the worst in the league at preventing chances against. As such, the point of this article will not be to restate a point that has already been made numerous times. Instead, I’ll be taking a look at Kane on the side of the rink at which he excels. Just about any hockey fan can confirm Kane’s offensive prowess, but many statistical models don’t reflect it in the same way that watching a Blackhawks game can. The wide discrepancy between Kane’s basic on-ice results and his analytical offensive impact led me to take a closer look into what the cause could be and what it could mean from a wider perspective.

Kane’s Play-Driving

The key offensive skill that Kane appears to lack is the concept of ‘play-driving’. Play-driving has gained a lot more traction in recent years, specifically in the analytics community, and has become one of the main methods of assessing a player’s offensive impact. The intent of play-driving is to quantify how well a player tilts the ice in his team’s favor. This concept is obviously inherently qualitative, so assumptions must be made. Play-driving is not a factual metric; it is instead a proxy, or a metric intended to represent something that cannot be easily quantified. The primary concrete statistic that makes up play-driving is shot attempts. As the thinking goes, if a team is consistently getting shot attempts at the offensive side of the rink with a certain player on the ice, then that player must be a good play-driver.

As a Canadiens fan myself, play-driving quickly endeared itself to me given how deeply ingrained it is into Montreal’s playing style. For multiple seasons now, Montreal has set the standard for tilting the ice in one’s favor. The Canadiens have consistently excelled in play-driving metrics such as xGF/60 and CF/60 thanks to a number of players who take an abundance of low-danger shots whenever the opportunity arises. However, this in itself is the potential flaw with play-driving: does creating low-danger scoring opportunities really correlate with actual goals for?

In Kane’s case, the answer is a resounding no. Here is a look at Kane’s RAPM chart from Evolving Hockey between 2018 and present-day.

The xGA/60 and CA/60 columns likely catch your eye immediately, but there’s no need to once again dive into Kane’s defensive ability or lack thereof. I’d like to direct you instead to the two columns prior - the xGF/60 and CF/60 columns - which are generally considered to represent play-driving ability. The two stats themselves are based on shot attempts, so they correlate with a player’s play-driving prowess. What is unusual about Kane’s chart is the fact that his less-than-stellar play-driving stats appear to have little to no effect on his GF/60 column, or actual goals. Despite being below average in both xGF/60 and CF/60, Kane remains outstanding in GF/60, at almost three standard deviations above average. His success in that column is supported by his outstanding scoring results; as previously mentioned, Kane has put up points at a rate almost unmatched league-wide over the course of his career, even into his early 30s.

The discrepancy between expected goals and actual goals is supported by other statistical models as well. While Kane is tied for third this year in 5v5 goals with 8, he ranks a measly 77th with 3.3 5v5 expected goals per Money Puck. The difference between the two, 4.7, ranks him second in goals above expected at 5v5. This disparity can be attributed in part to Kane’s shooting talent, which Money Puck rates at 27.4% above average this season. That being said, such a large discrepancy occurring so consistently must have more of an explanation behind it.

Playing Style

The dearth of scoring chances suggested by Kane’s lower expected goals totals contrasts with the high-end offensive opportunities he appears to create often by the eye test, something which I believe can be attributed to Kane’s unique playing style. I identified two major aspects of Kane’s game that could offer an explanation. For starters, Kane is a fantastic transition player. CJ Turtoro’s viz below (with manually-tracked data courtesy of Corey Sznajder) displays Kane’s proficiency in transition in percentiles from 2017-2020.

While excellence in transition is a noticeable asset while watching a game, it does not in itself contribute to shot attempts. Entering the opposing team’s zone, while useful, does not inherently result in scoring chances. Watching Kane transition up the ice certainly makes it appear as though he drives play well, but the primary metric behind statistical play-driving - shot attempts - is not affected. As a result, Kane’s transition proficiency does not aid his statistical play-driving profile.

The second, and likely more influential aspect of Kane’s game which contributes to the aforementioned discrepancy is his shot selection. A player like Kane chooses to be patient with the puck rather than repeatedly shooting from low-danger areas. On the other hand, someone like Montreal’s Brendan Gallagher constantly sends pucks to the front of the net, resulting in shot attempts and accumulation in expected goals. Here’s Gallagher’s RAPM chart since 2018.

As is clearly visible, Gallagher’s play-driving stats (xGF/60 and CF/60) are just about off the charts. It likely comes as a surprise that a player who scores so little relative to Kane is so much better by Evolving Hockey’s expected goals and Corsi models. To be clear, these results do not mean that Gallagher is better than Kane offensively. RAPM charts are only one resource out of many and should not be considered the utmost authority on individual player impact. The massive difference between Kane’s and Gallagher’s scoring totals should immediately be a red flag when comparing their play-driving stats. There is simply a flaw with how play-driving is calculated in many cases.

Former Toronto Marlies assistant coach Jack Han wrote a fantastic article on this subject. Patrick Kane, rather than surrendering possession by taking a shot that is unlikely to go in, prefers to hold onto the puck in hopes of creating a better scoring opportunity. Unfortunately for Kane's analytical profile, models don’t pick up on plays that don’t result in measurable events. As Han said, “Few models, statistical or otherwise, are built to handle events that don’t occur.”

In other words, because Kane’s effectiveness often comes from his willingness to wait for a better play to open up, he creates fewer events on which statistical models are built, like shot attempts. While Kane has been very successful with this method throughout his career, that success is not reflected in many analytical offensive metrics. It is exceedingly easy to see offensive patience and creativity while watching games, especially when it translates to goals as it does in Kane’s case. As a result, there is a disconnect between the analytics and the eye test.

Conclusion

So what does all of this mean in a broader context?

I want to emphasize that by no means am I discrediting analytics by exposing a flaw that occurs with a certain player archetype. That said, it is important to recognize these potential flaws and factor them into our evaluations. Patrick Bacon recently published an incredible article detailing the shortcomings involved with RAPM.

To me, Patrick Kane stands out as the perfect example of where the eye test and analytics diverge from one another. Plenty of models do take into account concrete scoring (like Evolving Hockey's Off stat), and Kane grades out quite nicely in those, so this article merely aims to discuss the flaws with models built on proxies.

While analytics remain the best tool we have to evaluate and compare players and teams, they still have a long way to go before we can trust them completely. In fact, it’s unlikely that we ever get to that point. A player like Kane shows perfectly why that’s the case. In a sport that is influenced by as many factors as hockey is, we must evaluate using every resource at our disposal in order to get as close to being right as possible. That includes our own eye test.

As I wrote this article, the unreliability of statistical play-driving became more and more apparent. I firmly believe that play-driving cannot be quantified - at least not well - because it often has a poor correlation with actual scoring results. After all, winning hockey games is about scoring more goals than the other team, so why use a proxy metric that doesn’t correlate with winning?

All stats are as of February 28, 2021