clock menu more-arrow no yes

Filed under:

The Phil Kessel Situation: When new metrics challenge what we thought we knew

New, comments

Arguing about Phil Kessel like it’s 2015 again.

Pittsburgh Penguins v New Jersey Devils Photo by Bruce Bennett/Getty Images

The Penguins are trying to trade Phil Kessel. They’ve made no secret of that. Even going so far as to leak that he that nixed a trade to the Minnesota Wild that would have netted the team Jason Zucker and swapped Jack Johnson’s bad contract for Victor Rask’s bad contract. Kessel can block trades to all but eight NHL teams and he invoked that right in this situation.

The reported trade sparked several different discussions among fans but the most interesting to me was the extreme variance in people’s opinions of Kessel as a player. He was a point per game player last season, which was good enough for 23rd in the NHL. That puts him among the very best scorers in the league.

But if we dig a little further into his numbers, he was worth less than half of a win according to WAR at Evolving Hockey. That’s because he took lots of penalties last year and he isn’t much of a play driver at even strength. In fact, he was one of the fifty worst defensive players in the entire NHL last season by expected goal impact and his offensive play driving didn’t do enough to compensate for that.

On the surface, Kessel seems like a one-dimensional scorer, a sort of Super Thomas Vanek. So seeing lots of smart people so eager to extol his on-ice value in the hours following the trade rumors might be confusing to someone who just started following hockey. If everything you knew about Phil Kessel was based only on his results using the best stats we have available, you might wonder what everyone was going on about.

But most hockey fans didn’t just start following hockey in the last year or two. And anyone who has been a fan for longer, particularly on Twitter, has been litigating Kessel’s career for well over five years.

The Phil Kessel Situation

Kessel, just by the nature of his playing style and the teams who have employed him, has been a constant source of arguments. When he was in Toronto, he became the media scapegoat for the failure of the team. That was unfair. The Leafs’ history was an overwhelming tome of page after page of failure. Kessel was just the latest character to take the fall for the sins of an antiquated and overmatched front office.

That alone would have led to pushback from a fanbase already tired of its team’s leadership and the media covering it. But the burgeoning use of better statistics exacerbated an already contentious conversation. The stodgy old media types wrote about Kessel being lazy defensively. They poo-pooed his scoring, saying the team couldn’t win with a one-dimensional player like him. They had no evidence for their claims other than their eyes, which fans already knew not to trust because those same eyes saw Randy Carlyle as a good coach.

Fans and public analysts pushed back on the narrative saying that the media, not Kessel, were the lazy ones. They used his scoring and his positive impact on shot differentials to dispute the arguments about his impact. Even if he’s not great defensively, as long as the offense compensates, then it doesn’t matter.

And that appeared to be true for much of his time in Toronto. The best available metrics at the time were relative shots. And when Kessel was on the ice, the Leafs were usually better than when he was off. A top-of-the-league scorer with break-even shot impacts is a valuable player and certainly not deserving of the blame on a team full of players who didn’t belong in the modern NHL.

So fans dug in their heels. Kessel wasn’t the problem and they did their best to counteract the media narratives. Reporters and writers also dug in their heels, although it’s difficult to discern whether they did so because they genuinely believed Kessel was the issue or because it got them clicks.

Ultimately, the Leafs shipped Kessel to Pittsburgh for a decent package of prospects while retaining some of his salary. The media shivved him one last time on the way out, most notably in an “article” by “noted” Toronto “writer” Steve Simmons in which the author bizarrely and incorrectly attacked Kessel for eating a hot dog on the way to every home game. Kessel responded by having the best year of his career and winning a Cup with his new team. Fatality.

But with Kessel on the trade block again, and for many of the same reasons as last time, we might have to reopen the book we all slammed shut emphatically as our disheveled anti-hero raised the cup over his head.

Revisiting the Phil Kessel Situation with better metrics

Using better stats than we had the last time he was available in the summer of 2015, Kessel’s most recent season looks rough. But this is only the second worst season of his career. What was the worst? Inconveniently for all of us who made up our minds about him years ago, his worst season was 2014-2015, just before the trade out of Toronto.

His two worst seasons have things in common. Primarily that he’s bad defensively. And with a broader view, he’s been bad defensively every season since 13-14 with the exception of his career year in 15-16. The only question has been how much offense he could generate to compensate for it. In some seasons, he’s been able to do enough to make himself a two-win player according to WAR. In some, he’s been closer to one. And twice, he’s hovered at or below replacement level. If Pittsburgh can successfully trade him this summer, both of those seasons will be followed immediately by a trade.

The Kessel case is an instructive one. Primarily, it shows how our evaluation of players changes over time. New metrics give us better measurements and better measurements mean we have to change our opinions to incorporate the new information. To try to visualize this, I created the following chart. It shows the difference between how we would have evaluated a player in the mid 2010s and how we would evaluate them now.

The data set includes all players who played significant minutes in at least three seasons between 07-08 and 16-17. The blue dot is where they were at the end of 16-17 using points and relative shot differentials. The orange triangle is where they would have been if we’d had WAR and xG impacts in those days. The line is the distance between the points and can be thought of as the difference between how we used to think about a player compared to now. The player highlighted here is Phil Kessel.

This helps show what we discussed above. We used to think Kessel was one of the best players in the league. Even though he wasn’t much of a play driver, he was one of the best scorers. But if we’d had today’s stats then, he would have looked much closer to average. He was slightly above average in WAR but slightly below average as a play driver.

This gap helps explain some of the conversations we see around him today. People who made up their minds after 16-17 and haven’t revisited those ideas still think of him as the blue dot. People who either weren’t around for all those old arguments or who’ve adopted the new metrics think of him as the orange triangle. Hence all the consternation.

What we can learn from the Phil Kessel Situation

But really, if we’re trying to be honest in our assessments of players, the correct thing to do is to use the new metrics. They’re better. And in some cases, like this one, that means admitting being wrong about things.

I overrated Kessel based on the information at the time. I have to admit that with new information, the idea that his defense was so bad that it detracted from his overall value in a meaningful way is true. That doesn’t mean all of the clowns who were spouting off about it at the time are suddenly geniuses. Blind squirrels and nuts, etc. But if I don’t admit I’m wrong when I’m wrong, then I’m one step closer to writing a poorly informed post about Tyler Johnson eating a big bag of popcorn on the bench while everyone else does skating drills in practice every day.

One of the fun things about being relatively early in the analytics movement in hockey is that things are changing quickly. Most of the tools we used to evaluate players three years ago have already been replaced with better ones. That will continue for the foreseeable future even as the pace of those changes decreases.

The best analysts will adapt. Those who don’t will be left behind becoming more and more outdated with every new tool they fail to adopt. People still stuck to opinions they formed in 2015 already sound silly. In another five years, those opinions will be laughable. And five years after that, they’ll be irrelevant.

That’s the choice for analysts. Either accept new information that will often prove your previous opinions wrong or become irrelevant.

Some Other Stuff

If you want to see how other players look in the chart used in the article, you can generate them in the app here. I recommend using it on desktop only as it hasn’t been optimized for mobile.

The chart includes all players who played significant minutes in at least three seasons between 07-08 and 16-17. The players in the drop-down are sorted by the biggest change in evaluation so the players at the top of the list will be the most interesting to look at. The goal of the tool is to help see which players we might be holding on to incorrect opinions about if we haven’t revisited their performance using new metrics.

All data for the app and this article is via Evolving Hockey. If you can, consider supporting their Patreon here.