4

By The Numbers: Jets strengths and weaknesses in each skater

Due to my day job working with a data tracking and consulting company, I am unfortunately handcuffed in the amount, extent, and type of analysis I can provide through public forums. That does not mean that I am unable to provide any type of analysis. I am limited to using other individual’s ideas and models rather than my own. This does allow me to highlight the good work of others.

One such individual is that of DTMAboutHeart and his WAR model. I get that complex models can be difficult to understand the theory behind them or suss out the information within.

So instead of just using a model as evidence with studying the Jets, I will use the Jets as an example of how and why you would use such a model and what it means.

What is WAR?

No, I am not asking what it is good for.

WAR stands for Wins Above Replacement. It translates multiple different output statistics into one currency, allowing you to combine these together. Thus, one can approximate a player’s overall impact on game outcomes, or wins. This currency is given relative to that of an easily replaceable player.

DTMAH’s model uses many well known statistics, like point production, quality of teammate, shot differentials, and shot quality variables, to estimate a player’s overall impact on the team’s goal differential, which in turn impacts the win column.

This WAR model could be broken down into three major components:

OBPM – Offensive Boxscore Plus-Minus

If you want the ultra-simplified version, OBPM really is a player’s impact through their individual offensive contributions.

The model looks at goals, primary assists, secondary assists, individual shot volume, individual shot quality, and other factors, all adjusted by linemate quality over the season.

OXPM – Offensive Expected Plus-Minus

If you want the ultra-simplified version, OXPM really is a player’s impact through their ability to positively impact their linemate’s offense.

The model looks at a player’s impact on expected goals for, a shot metric containing both shot quantity and quality aspects, and adjusts at the play-by-play level according to usage like linemates, linematching, zone starts, head coach on both teams, and schedule.

DXPM – Defensive Expected Plus-Minus

If you want the ultra-simplified version, DXPM really is a player’s impact through their ability to negatively impact their opposition’s offense.

The model looks at a player’s impact on expected goals against, a shot metric containing both shot quantity and quality aspects, and adjusts at the play-by-play level according to usage like linemates, linematching, zone starts, head coach on both teams, and schedule.

If you would like a more detailed series on DTMAH’s WAR model, please click here.

How do you use it?

WAR is an all encompassing stat, but it should never be used as the end of analysis. Rather, one should use WAR to start the conversation or analysis, and then be broken down further to gain a better understanding of what’s going on. Start with WAR, and then move on to the other models, statistics, and methods used by analysts.

WAR allows you to see the estimated overall value of a player. It also allows you to distinctly see where a player helps and hurts the team. However, WAR cannot tell you why it is correct or off. It cannot tell you if one should expect the player to regress or not. It cannot tell you how to improve a player. It cannot tell you which player will fit into a team’s system better.

From Fangraphs:

Given the nature of the calculation and potential measurement errors, WAR should be used as a guide for separating groups of players and not as a precise estimate. For example, a player that has been worth 2.4 WAR and a player that has been worth 2.1 WAR over the course of a season cannot be distinguished from one another using WAR. It is simply too close for this particular tool to tell them apart. WAR can tell you that these two players are likely about equal in value, but you need to dig deeper to separate them.

However, a 2.4 WAR player and a 0.5 WAR player are different enough that you can have a high level of confidence that the first player has been more valuable to their team over the given season.

First comes WAR, then comes the three components, and then comes the other analysis (like Corsi, P/60, WOWYs, etc.).

Jets Defenders

OBPM, OXPM, and DXPM are per 60 minutes and on the left y-axis, while GAR is overall value and on the right y-axis. Note: The values are for even strength situations only.

The model suggests that Jacob Trouba is the Jets’ best defender in overall impact, with Ben Chiarot being the weakest defender.

Trouba’s individual offense is tied with Paul Postma for the highest OBPM rating on the Jets and ninth highest rating, with Florida’s Alex Petrovic, for all defenders. Byfuglien comes right after those two, with Toby Enstrom having the weakest impact.

This should make sense with Postma and Trouba ranking 9th and 11th for 5-on-5 point production relative to ice time, and Byfuglien at 15th. Meanwhile Enstrom falls just 7 spots from the bottom.

While 5-on-5 production is not the only way a player impacts OBPM, it is a major driver and helps us suss out why players fall where they are.

Going one step deeper, we can look at a player’s historical impact or look at shooting percentage regression to determine the sustainability of a player’s OBPM. For example, Postma has never scored remotely at this level before or posted nearly as strong of an OBPM. It is likely that OBPM overstates Postma’s true-talent level, although by how much remains to be seen.

OXPM looks at how a player improves team offense through tilting the ice and allowing the team to generate more chances. It suggests Trouba (2) and Byfuglien (10) are both top 10 defenders in this area. They allow the Jets opportunities that raises the Jets expected goal production. However, DXPM shows how Byfuglien’s offense comes at cost with his below replacement level defensive impact in expected goals against.

Overall the trade off is still highly positive. Dustin Byfuglien ranks 45th overall in overall impact for NHL defenders. He is still a highly positive impact defender despite his fall from his typical top-10 defender impact.

DXPM also shows us that, while Enstrom has struggled offensively, the small defender does still provide value with being the Jets best defensive player after Trouba.

One potential issue with the XPM statistics is that they use priors, or historical performance to regress this season’s performance. A positive to using priors is that you worry a lot less about inflated values from outliers like we discussed with OBPM. A negative to using priors is that you worry more about when a player started with an inflated value or has no history.

Jets have an example of each.

Josh Morrissey as a rookie has no history, so XPM may be slow in reaching his “true talent level.” While Morrissey has been given the benefit of the Jets top players in usage, he also has been the Jets’ best defender in adjusted-Corsi%. This does not excuse his results, with other rookie defenders ranking higher than Morrissey, but it may still be slightly deflated.

Ben Chiarot started his NHL career at the same time as Byfuglien swapped from forward to defense during the 2014-2015 season, and the two rarely separated that year afterwards. It has been well established that Byfuglien is a far superior defender than forward. XPM likely falsely assumed Byfuglien’s improvement was due to Chiarot joining with him. This is why we’ve seen Chiarot’s XPM values descending greatly with each season, but I question if it has fully reached the proper value since he’s had such a staunchly negative relative adjusted-Corsi% over the past two seasons.

Jets Forwards

OBPM, OXPM, and DXPM are per 60 minutes and on the left y-axis, while GAR is overall value and on the right y-axis. Note: The values are for even strength situations only.

The Jets forwards are interesting due to there being a distinct split in how each player contributes to the team offensively.

We see here that Mark Scheifele, Nikolaj Ehlers, and Patrik Laine have all excelled in terms of their individual outputs, like scoring, but trail in terms of improving the Jets expected goal generation by tilting the ice.

Meanwhile, veterans Blake Wheeler and Mathieu Perreault are the complete opposite, with both struggling to score but have been a positive influence in driving expected goal generation through improving their linemates’ shot volume and quality.

Bryan Little has been more of a mixed player offensive, both producing and improving their linemates offensively.

Interestingly, Adam Lowry has been a below replacement player offensively in his individual offensive outputs but his OXPM value is extremely high, and far higher than his historical norm. I am curious to see whether or not his OXPM value falters next season. He is still legitimately one of the Jets better defensive player.

Speaking of defensive value, Joel Armia seems to be the Jets best defensive impact players after Perreault. The only issue with Armia is that the offensive side of things still lags substantially behind his defensive game relative to Perreault.

This graph aptly displays the issue with the Jets’ bottom-six. The Armia-Lowry-Matthias line lack sufficient scoring punch for the minutes they handle. While the trio together are able to match up against toughs fairly effectively in terms of reducing opponent offense with the defensive skills of Lowry and Armia, there just is not enough secondary scoring punch.

This is why I hope the Jets can develop a secondary scoring line next year between Nic Petan, Marko Dano, Jack Roslovic, and Kyle Connor next year. This would slot the three of Armia, Lowry, Shawn Matthias, and Andrew Copp into a more suitable role of an elite fourth-line on a potentially contender team.

All numbers are courtesy of Corsica.hockey unless otherwise noted.


MORE FROM GARRET

  • Joe hannah

    Interesting article, but what I find interesting is this idea that a Connor and Roslovic and Petan can make a good line. The moose have tried putting Connor and Roslovic and except on the power play it does not work – every time they are put together on 5 on 5 they are a high minus. The reason is non of the above players want to go into the corner and fight for the puck and they can’t showcase their skills fighting on the boards which is why Connor is seeing success with Cormier and Lipon. Unless those 3 are in the top 6 they will not play together. Dano might be able to provide some of that. Petan is a total different story, he is skilled and stick handle like crazy but this is not junior and he is just not as effective as one would have hoped in the NHL and frankly I am not convinced he will be. He is one of those guys who is not good enough to be top 6 but is to skilled for bottom 6.

  • ground control

    It is hard to know how Nic Petan fits on this team. But from what I can see there is the NP wasting on the fourth line, and there is the NP that is totally at ease on the power play. I would like to see him on a third line with skilled players to see what he could do.