clock menu more-arrow no yes mobile

Filed under:

Expectation-Beater Kyle Palmieri & The Devils’ Actual Goals Against Expected Goals

More of the 2017-18 Devils skaters beat the expected goals model at Moneypuck than the prior two seasons, but Kyle Palmieri beat the model in three straight seasons. This post goes into how the Devils did against the model and what that could mean.

Tampa Bay Lightning v New Jersey Devils - Game Four
Kyle Palmieri just exceeded the expected goals model in three seasons per Moneypuck’s model.
Photo by Bruce Bennett/Getty Images

Not that long ago, CJ made the case that the 2017-18 New Jersey Devils were not as lucky as one may think. It’s a fair set of arguments based on various stats. However, there is one aspect that was definitely favorable for the 2017-18 Devils compared with previous seasons: individual scoring against expectations.

One of the newer concepts in hockey analytics is expected goals. It is a model that weighs various aspects of an individual shot taken by a player and determines, based on past data, the probability of that shot becoming a goal. This sort of model has been used to determine whether a player’s actions on the ice should have resulted in a more or fewer goals for or against than what actually happened. It can also be used as any to identify whether someone was fortunate or not when it comes to goal scoring.

There are a couple out there and the one I will be using today comes from Peter Tanner’s Moneypuck. It is a bit different from the one used at Corsica, a place that claims that 2017-18 NHL MVP and Superstar Taylor Hall only scored 38 goals in all situations last season. The logic is the same. The Moneypuck About section there has an explanation of how it all works and the stats page for players has a clean and quick layout that is accurate and easy to understand. If there is a criticism, then it is that it does not differentiate between teams for players that move within a season. For example, the numbers for Patrick Maroon include his time with Edmonton and New Jersey - there’s no way I could tell on how to break it up.

The data at Moneypuck goes back three years and so I took last season’s roster, pulled their actual goal scoring, and found that the 2017-18 Devils had more players beat the model than the previous two seasons. In all situations, the 2017-18 Devils had nine players beat the model by at least one goal compared with only three and seven respectively in the last two seasons. The 2017-18 Devils also had only six players fall below their expected goals count by more than one goal; that is an improvement over the 13 and 11 players who were also below the model by more than one. That’s all good to see. One could question whether it is “luck” that more players were within a goal or exceeded the expected goal count by more than a goal compared with the past two squads. I think favorable is the appropriate adjective.

What you should also takeaway is how great Kyle Palmieri has been compared with the model.

Goal Scoring in 5-on-5 Situations

Here is how the goals scored by the 2017-18 skaters (with at least 100 minutes played) stacked up against the model in 5-on-5 play. Players in grey are no longer with the team. Green numbers are for seasons where the expected goal model was exceeded by two or more; red numbers are for seasons where the actual scoring was less than the model by two or more.

2017-18 Devils roster Actual 5-on-5 Goals vs. Expected Goals
2017-18 Devils roster Actual 5-on-5 Goals vs. Expected Goals
Data from

I went back to 2015-16 to show any changes from prior seasons. You can see how much Ray Shero has changed the roster in a relatively short amount of time. Only five skaters played enough in the last three seasons to appear here, and that number is down to four since John Moore signed with Boston.

Anyway, Palmieri is the only Devil to have beaten the model by at least two goals in each of the last three seasons. Note that it is not for a small amount of goals either. Palmieri has been a consistent 10+ goal scorer in 5-on-5 play. While his shot selection and distance suggest he should not have been this productive, he absolutely has been. For the 2018-19 Devils to thrive, the hope is that Palmieri can stay healthy enough and keep on firing ahead of expectations. With three straight seasons of doing so, I’m expecting him to beat the expected goals model again.

As for the rest of the 2017-18 roster, you can see some surprising gains. Damon Severson beat expectations more than anyone else, which is a big gain after two straight seasons of not meeting them. Miles Wood went from being below the model by nearly four goals to exceeding it by nearly four goals in a season. In 2016-17, the expected goal model would suggest that Taylor Hall was rather unfortunate that season. Things were set to right in 2017-18 as Hall scored right around where the model thought he should in 5-on-5 play.

It was not all rainbows and sunshine, though. The now-Bruin Moore went from beating expectations big-time to being right around them between 2016-17 and 2018-19. I could see how some may interpret that to be a cautionary occurrence with respect to Severson. Stefan Noesen beat expectations to get to eight goals; he finished right around the model in 2017-18, albeit with more goals scored. That should put a bit of a damper on, say, expecting Wood and Severson from exceeding the expected goals model. Only Palmieri and Steve Santini actually beat the expected goals model in consecutive seasons over the last three Devils seasons - and Santini did so by scoring all of two goals.

On the opposite end, three very different Devils failed to meet the expected goals model by more than two goals: Blake Coleman, Pavel Zacha, and Drew Stafford. Stafford not producing as much as the model expected suggests that his usefulness was not as limited as it seemed. That said, if the role was for him to chip in goals and take shootouts, well, not actually producing as much kind of justifies not immediately bringing him back. I recall Coleman being real snakebit at times last season. He got hot in March with some slick goals out of seven markers; he had six in the rest of the season. Still, the model thought he should have had more. At least Coleman was killing clock and frustrating opponents on the PK and in 5-on-5 play. Zacha is an odder situation since he also missed the expected goals model by three last season. Your mileage may vary if Zacha had scored seven more goals across two seasons would improve his standing on the roster or among the fans. Since he is a young player and it is not apparent what the Devils exactly have planned for him, I want to give him the benefit of the doubt. That said, not meeting the model twice in a row and by at least three goals suggests he’s rather unlucky - or he needs to work on his shooting in some way.

Goal Scoring in All Situations

Goals are goals in all situational play. The model can incorporate special team play, which is crucial from a goal-scoring perspective given how some players have been real monsters on the PP. Here is how the goals scored by the 2017-18 skaters (with at least 100 minutes played) stacked up against the model in all game situations. Players in grey are no longer with the Devils.

2017-18 Devils roster Actual All Situation Goals vs. Expected Goals
2017-18 Devils roster Actual All Situation Goals vs. Expected Goals
Data from

Taylor Hall went from being right around the model in 5-on-5 to soaring past it when you include special teams. A delta of 8.2 was one of the highest in the NHL last season; only 18 skaters had superior differences between actual and expected goals. It is a remarkable achievement considering how the model thought he should have scored several more goals in 2016-17.

Yet, the real highlight is Palmieri. Once again, he beat it in all three seasons. Only Moore and Santini beat the model in consecutive seasons, but neither resulted in the sheer amount of production Santini had. When adding the power play, Palmieri just crushed the model. If only he was able to play around 80 games last season; he could have had a chance at 30 goals again. While other failing to repeat make me think this may not be so repeatable; I am not doubting Palmieri in beating expectations for a fourth season.

Adding all situations causes some notable shifts. Sure, Severson and Wood remain high up with respect to beating expectations; but Jesper Bratt and his very good one-half of last season is moved up here. While not in that significant range, Brian Boyle is in the positive half. Moore isn’t a Devil anymore, but those three additional goals in all situations had him exceed the model compared to just being ahead of it in 5-on-5 play. Travis Zajac got nudged onto being above expectations by a bit compared to just meeting it in 5-on-5 play. It is also his first positive comparison against the model after two straight seasons of missing it by about a goal. Not all of them were positive. Nico Hischier got bumped down to the negative half of the list; which speaks to how he did not really succeed on special teams. I think once he does get a more significant role, he will be productive. It is a concern on how to fit him in; looking at this chart, I would not touch Hall or Palmieri and assuming the coaches stick with a 1-3-1 formation with one defenseman, that’s not a lot of options for Hischier. We’ll see though.

Looking at all situations did not help out the deltas for Stafford, Coleman, and Zacha. Stafford did have some special teams work at points, so he was able to add some more goals - but not enough of them per the model. Coleman did not receive a lot of power play time, so that difference of four goals could be through shorthanded goals. After all, one-on-ones, two-on-ones, and rush plays in shorthanded situations are desirable plays - and Coleman did not immediately hit on them in 2017-18. If he’s able to replicate those situations, then maybe he gets some more puck luck, and gets to contribute even more than just making a shorthanded chance happen. Zacha also remains low; but note his 2016-17 season. He did not miss the model by much then. That provides some hope for the future; although, it would be ideal for Zacha to make the most of his chances.

So What Does this Mean?

I agree with much of what CJ wrote about how the 2017-18 was not so lucky. I will say that looking at expected goals vs. actual goals for the individual Devils players from that season suggests there was some favorable events. However, the larger gains across the roster also adds further to the real notion that the 2017-18 Devils were improved compared to the 2015-16 and 2016-17 seasons. Not that you, the reader of this fine website, need much convincing of that; but it also gives me a bit more confidence that they will probably not fall back to that level. That said, it does make me wonder whether we’ll see the same amount in 2018-19. After all, only Palmieri, Santini, and Moore were able to beat the model in consecutive seasons in either (or both) 5-on-5 and all situation play. It is something to keep track of.

The more positive takeaway is that this should be seen as evidence that Kyle Palmieri really is a goal scorer. While he may not be among the best in the whole NHL, his big contribution is lighting that lamp. He has a strong shot, a strong one-timer, and he is unafraid to use it. On a team that could use more shooters and seemingly always more production, Palmieri has been and is a core player for the Devils. Again, I hope he’s able to play more than 62 games in 2018-19 and continue to exceed the expected goal model for his shots because his scoring prowess will help the Devils strive for a playoff spot for a second straight playoff spot.

What do you make of the Moneypuck expected goal model for the Devils? Are you surprised in a good way that Palmieri has done so well against it? Who do you think will turn it around for 2018-19? Please leave your answers and other thoughts about goal scoring and expected goals in the comments. Thank you for reading.