Revisiting Expected Goals for the 2018-19 New Jersey Devils

The concept of expected goals is a strong one. Based on historical data and primarily driven by a shot’s location and type, an expected goals model determines the value of any given shot with respect to how many goals it is worth. It is not an exact value or a truism; it is a model. Among other things, it does provide at least an idea of whether a player is doing the right things regardless of whether the shot is going in the net. Moreso than shooting percentage.

Back in December, I looked at how the New Jersey Devils were producing with respect to the expected goals model calculated by Sean Tierney. This model was and is for all situational play and all players were considered. Tierney’s calculation of expected goals (xG) is not the only one - Offsides Review has it, Natural Stat Trick recently added it - but it is a good source . His shot maps tableau also presents where the player’s shots were recorded by the game’s scorer. It is an effective graphical representation that supports his results. Since the 2018-19 season is now mercifully over, let’s revisit Tierney’s tableau and see how well the Devils players met the expected goals model. Thanks to Tierney for maintaining his public tableau; please check it out and consider following him on Twitter, @ChartingHockey.

The Team xG - Something that Went Well and Something that Did Not

Before jumping into the players, Tierney’s model provides a little bit of a silver lining for the 2018-19 Devils. In all situations and based on their Fenwick*, the model had the Devils scoring 217.6 goals. The Devils scored 219. That is not a huge difference. However, it is something where the Devils were just a little bit ahead of expectations. The Devils finished 29th last season so any small positives should be highlighted.

Unfortunately, this silver lining had another, darker side. A team expected goals value of 217.6 is quite low. As is an actual goal total of 219. Both rated among the bottom ten teams in the league. What this means is that while the Devils were slightly better than expectations, they could have done a lot better in general on offense. Their relatively low generation of shots and shots from dangerous locations suggested that they would not score a lot of goals. They pretty much did that in 2018-19. Those lamenting about the team’s lack of talent, especially up front, are justified in part by the expected goals model and reality. That near two-month period where more of Devils’ forwards were from Binghamton than initially with New Jersey hurt.

*Note: What Tierney calls shots is actually unblocked attempts - shots on net and misses. This is known as Fenwick. I regret not picking up on it in the original post in December. That’s why you’ll see it here.

The Four Departed Devils of 2018-19 by xG - They Met It At Least

The Devils lost four players last season with three being dealt away prior to the 2019 NHL Trade Deadline. Teams that are sellers by the deadline tend to get worse and the New Jersey Devils were no different when Brian Boyle, Marcus Johansson, and Ben Lovejoy were sent to Nashville, Boston, and Dallas, respectively for draft picks and Connor Carrick. The Devils also lost Jean-Sebastien Dea to waivers earlier in the season, which had an impact. By the way, no one can actually score a decimal of a goal, so rounding up the model’s results is encouraged.

Sean Tierney

These four Devils were responsible for 29 goals, or 13.2% of all goals, last season. It is not a huge amount on its own, but it is not nothing. I do not even think it was really replaced either. What’s important to note is that by how and where they were shooting the puck, the expected goals model aligned with their results. If anything, this foursome out-performed the model by a little bit. It is not the only way to evaluate a player or their performances. However, it is evidence that their production should not be seen as a total surprise based on what they did on the ice. They did not underachieve purely from a goal-scoring perspective. They were missed to a degree. Especially Boyle and Johansson when the Devils were forced to go five-plus forwards into their depth chart to have a full roster.

It is also of note that these four also put a good percentage of their Fenwick on target. Johansson and Dea putting up over 77% of their unblocked attempts on net was quite high on the Devils last season. Among players who played at least 200 minutes last season, only Joey Anderson (83.05%), Brett Seney (80.33%), and Sami Vatanen (78.1%) were more accurate than those two. Granted, only Johansson and Vatanen had over a hundred unblocked attempts last season so it is not like that group had a lot of shots on net anyway.

The 2018-19 Devils Defensemen by xG - Damon Severson’s Standing Out

The Devils saw twelve different defensemen put on the bedeviled NJ last season. Three were definitely of the “call them up and see what they can do” variety. Four did not even play 200 minutes with the team so I greyed out their names. I would take their results with a grain of salt. Especially since none of them scored a goal. The other eight were used more often as they took regular shifts when they did play.

Sean Tierney

Damon Severson is in green because he exceeded the expected goals model by at least two. It was nearly three. That does not seem like much but look at the rest of the blueline. Nobody else on the Devils even scored six goals in all situations. Severson may have his critics but his production was a positive. And while he was not as accurate as Vatanen, an on-target percentage of 76.84% is ahead of most of the defensemen. Seeing that a lot of defenders are shooting from 45-55 feet away, I would expect there to be plenty of missed shots. But Severson’s aim and finishing was true in 2018-19. Hopefully he can keep it up.

But here is the thing. Based on the model, the other defensemen scored about as many goals as expected. If they were off from the model, it was by less than one - which is not really underachiveng. This points to an issue of defensemen just not taking dangerous enough shots. Given how they are used and where they set up on offense, this is not a total surprise. Long shots have an inherently low xG value assuming the shot even gets to the net through bodies and/or sticks. However, given how often the Devils defensemen pinched last season and how often they “activate” and join the attack, there just was not a lot being generated - and so not a lot was.

And this is very much true for the forwards; the players who do play closer to the net and tend to take more scoring chances and create threatening opportunities to score.

The 2018-19 Devils Forward by xG - Kyle Palmieri Crushed It, Stefan Noesen Crushed By It

Here is the big, long list of all of the forwards that played for New Jersey last season organized by expected goals. Players in green exceeded their xG by two or more. Players in red scored two or fewer goals compared to their xG. Players that are greyed out played less than two hundred minutes so their results should be taken with a grain of salt. That group impressively includes Kurtis Gabriel seeing that he played in 22 games. And I cannot argue that he deserved more ice time anyway.

Sean Tierney

Back in mid-December, I noted how Kyle Palmieri, Travis Zajac, Brian Boyle, and Jesper Bratt exceeded their respective xGs. Even back then, Palmieri was the run-away leader in both goals and beating the expected goal model. The model figured Palmieri’s shots would yield about 8.7 goals and he had 16 then. While the rest of the season did not lead to Palmieri scoring 30+, he still led the Devils in goals and in smashing the model’s results. The gap even increased a little bit; Palmieri’s 27 goals out performed the model by 7.43 goals. The Pride of Montvale, New Jersey may not be pleased with how the team performed this season; but his production was not at all a problem.

Zajac also continued to beat the expected goals model by a healthy margin. Among the non-greyed out players, Zajac had the highest shooting percentage among forwards. With his overtime winner in Florida, Zajac’s 19 goals is the most in a single season for him since the 2009-10 season when he scored 25. Zajac’s stick was hotter than usual and that helped him beat the model by nearly four goals. His off the puck play and on-ice rates suggest that he is not what he used to be, but as with Palmieri, Zajac’s production was far from a problem.

Jesper Bratt barely made the green cut back in December and he still stayed there. He did not beat the model by a lot, but he had a couple more (2) than the model suggested. Bratt could stand to create and take more unblocked attempts himself. His injuries interrupted his season and that was a factor as well. But it is encouraging for the future. Pavel Zacha and his late run moved his totals past his xG by nearly three goals. I echo Mike’s post that we should be cautious as to whether Zacha has turned any kind of corner. However, even if he did not get hot, an xG of 10.43 was one of the better ones among forwards on the team in 2018-19. That suggests that Zacha was putting in some positive work on offense. Of course, as with the rest of the team, the team really does need to raise the bar.

Moving on to the players that did not exceed or miss the model by a large margin, I want to note a few of them. First, Taylor Hall went from not meeting the model in mid-December to just surpassing it before his injury. Had Hall kept on playing, I would like to think he would have one of the highest xGs on the team. Had he maintained an xG per game rate of 0.33, he would finished the season with an xG above 27. Alas, his knee would not let him do it. Second, the expected goals model suggests you should give Blake Coleman a little more credit. Instead of marveling that the winger scored 22 goals, you should be more or less satisfied. Coleman put in a lot of work to generate nearly as many unblocked attempts as a top-six winger like Palmieri. His work was rewarded and it could even be argued he was short by one. Of course, will Coleman have the same opportunities next season? Probably not. But in the spots he was given, he produced. Third, after Coleman, Hall, and Hischier, the rest of the forwards who more or less met the model provided a small number of goals. As notable that Anderson and Seney were accurate, they scored just a bunch. As close as Kevin Rooney shot the puck, he gave the team only six goals. As they may have contributed in a few nice spots, only Anderson and Seney are young enough to think they could really improve. As I stated earlier, the Devils need to raise expectations to provide more offense in the future. This may mean getting more of that desired talent. Good luck, Ray Shero.

Three players stood out in a bad way by missing the expected goals model by a lot. Let’s touch on them.

Miles Wood has had his struggles in 2018-19 between injuries and playing his way down to a fourth line. He salvaged some of it at the end of last season to get to ten goals. However, he still missed the model by four goals. I focused on Wood’s shooting in particular back in mid-December as part of this post. I hope his shooting is more fortunate next season and he makes a point of it to be more judicious when he attempts to shoot the puck. He had no issue with generating unblocked attempts. The issue is that he rarely saw a shot he did not like.

Stefan Noesen was one of the unluckiest Devils in 2018-19. Down the stretch, Stefan Noesen would have one or two plays where he is in a great position to shoot and score and he just did not. He would miss. He would get stuffed. He would get robbed by the goalie or the cruel metallic goal frame. Not that Noesen is the offensive machine or this vital part of the offense. However, a low-scoring team cannot beg to be choosy for their scorers. Noesen’s work on the ice yielded an xG of eight goals. His average shot distance was one of the shortest on the team. If you adjust for ice time, his xG per 60 minutes was one of the highest on the Devils. He could have exceeded that by a little bit and provide ten, which is a good return for a bottom-six winger that primarily plays only at even strength. Noesen actually scored just three with a shooting percentage of just 3.33%. My point is that Noesen’s finishing betrayed him. He was on the cusp of a solid season of contributing - even after he returned from injury - but the pucks would not go in.

Michael McLeod did not impress a lot of fans when he got a real chance to perform in New Jersey in February and March of this year. Similar to Noesen, McLeod had done enough to at least score a couple of goals. His xG of 2.2 supports that. Similar to Noesen, McLeod was denied in a couple of key moments of that goal. This is amid plenty of shifts and games where McLeod did not make much of an impact. It is surprising to me to find that he appeared in 21 games. It is not surprising to me that he did not score a single goal last season and the expected goals model suggests that he should have had two or three. Hopefully it will be onwards and upwards for the Devils’ 2016 first round draft pick in 2019-20.

In total, Palmieri, Zajac, Zacha, and Bratt beat the model with ex-Devil Boyle falling just below two goals by the time he was dealt. Hall pulled himself out of the red; Noesen and Wood stayed in it, and McLeod joined them.

Conclusions & Your Take

The expected goals model is just that - a model. And depending on how it is calculated, it varies a little bit from site to site. Still, it is a good mechanism to get an understanding on how valuable a player’s shooting has been and compare it to how much they actually scored.

For the 2018-19 Devils, the team as a whole beat Sean Tierney’s xG model by a little bit. Five Devils beat the model by at least two goals: Kyle Palmieri, Travis Zajac, Damon Severson, Pavel Zacha, and Jesper Bratt. Three were behind the model by two goals: Miles Wood, Stefan Noesen, and Michael McLeod. On the surface,this seems OK. Good, even. It definitely is for the likes of Palmieri and Zajac, who beat the model by over four goals each.

However, the xG model by Tierney (and elsewhere) also point to how the Devils’ offense was lackluster last season. Yes, the team beat the model by a little bit but the model had the Devils at a fairly low bar to clear. Plenty of the players that met or slightly beat expectations scored fewer than ten goals each. Relative to the rest of the league, the Devils were not generating a lot of unblocked shot attempts (what Tierney calls shots in his tableau), much less scoring chances that would have raised expectations and possibly led to more goals. Those who felt that the Devils were lacking in offensive talent before, during, or after the rash of injuries at forward this season are justified to think this way due to the expected goals model. For a better 2019-20 (and beyond), the Devils will need to obtain more talent and increase their own chances by xG in the hopes of getting more actual Gs.

Once again, go check out Sean Tierney’s work. This is his shot map tableau and it is fun to dive in and see a graphical representation and a data summary of a player’s shots now that the season is over. Maybe you may find something that someone should continue or something that should make you wary about a player being signed and/or acquired. If nothing else, it is way to help pass the time in this current offseason.

In the meantime, what do you make of all of this? Are you surprised that the team has scored a little better than what the expected goal model suggests? How great is it that Palmieri beat the model by a lot? What about the others: Severson, Zajac, Bratt, and/or Zacha? Will Wood, Noesen, and McLeod turn it around for next season? What should the Devils do to increase their xG for next season? Please leave your answers and other thoughts in the comments. Thank you for reading.