By The Numbers: An Expected Goals look under the hood 33 games in

Updated: January 11, 2018 at 1:12 am by Garret Hohl

The season is thirty-three games in, so we take a look at the Jets and see what is making the team tick and perform the way they are.

In this article, we will use DTMAboutHeart’s Expected Goal model to see how the team performed for 5v5 situations.

A Quick Expected Goal Explanation

Expected goals is a term for any model that tries to describe how many goals we expected a team or player to score given some variables. This is why there have been many, many models with the same name.

DTMAboutHeart’s specific model is the most powerful one we currently have. In fact, the model predicts future performance relatively better than other shot metrics, like Corsi.

The model is a shot metric, meaning it adds up all shots, just like Corsi. What separates it from Corsi is that not every shot is viewed as equal. Each shot is weighted by its probability that a goal will arise from the shot, using variables like:


In addition, the model also looks at shooter history. It takes into account a skater’s regressed shooting percentage history, and then compares that shooter relative to league average. This also means any rookies without history would be considered a league average finisher, or those with very little history will be considered quite close to league average.

Team Level

Screen Shot 2016-12-16 at 10.26.25 AM

The blue dots are expected goal share and the red dots are actual goal share. The difference between the team and their dots are then created by goaltending talent, “luck” (variance), and potentially other factors not recognized by the model.

As it currently stands, the Winnipeg Jets sit 9th for points and 10th for points per game in the Western Conference. However, the team sits 8th in expected goal share.

The expected goal model suggests that the Jets may be ever-so-slightly better than their current record. Now, some of this difference between standings and expected goals may be a legitimate cause, due to skill factors like special teams and goaltending.

Long story short, the Jets are pretty much legitimately right where they are and deservedly so. The team will need to make significant changes, whether that is in player development of their young players, roster changes through trades or health, systems change to optimize the team more, or simply pulling together as a team and stepping up their game in order to improve.


Screen Shot 2016-12-16 at 10.09.14 AM

We can also use expected goals to determine how a player’s 5v5 goal scoring may be over or under performing what we should suspect given that shooter’s shot selections and their individual history.

Currently, the biggest under-performers are as most would suspect. Nikolaj Ehlers is about two full goals below what we’d typically expect from him, while Marko Dano and Matthieu Perreault are not far behind with 1.8 and 1.6 goals below their expected totals. None of Nic Petan, Joel Armia, or Alex Burmistrov have played in a while, but they are all just about due a goal.

The over performers are a lot more interesting. Mark Scheifele has scored a full 3.4 goals more than we would expect him to score on average, given his shot selection and his history. Blake Wheeler, Andrew Copp, and Shawn Matthias are all just under one goal over their expected performance.

Then we come to Patrik Laine. Laine is an interesting case. As a rookie, the model assumes Laine is a league average finisher relative to his shot selction placement, as it has no NHL history to go off of. However, prior to the season’s start we expected him to be an above average finisher.

So how much, if any, is over performance and how much is talent?

Well, as far back as we can go with this type of data, Steven Stamkos has been the best pure finisher to play hockey. Typically speaking, Stamkos scores about 1.5 times the number of goals one would expect given his shot choices. This means that if Laine were equally as dominant his “true-talent” expected goal totals would shift up from 3.84 to 5.76.

Of course there is still the chance that Laine is the best pure finisher in the NHL since the 2007-08 season, even more so than Stamkos. It is realistically possible, although not most probable. However, the improvement would be marginal. Currently Laine has scored 2.6 times more goals than the league average finisher would typically score given the shot choices he has taken.

This is far more than what we can reasonably expect even for a generational talent. There is no way Laine is 2.6 times the finisher as league average when the next best individual in modern history is only 1.5.

However, we can say that the perception of Laine being an elite finisher has stood up thus far, even if he has been over performing at 5v5.


Screen Shot 2016-12-16 at 10.11.47 AM

The Jets defenders are a lot simpler to go over.

Dustin Byfuglien and Josh Morrissey have scored about as many goals as we would expect, given their shot location and history. Morrissey, like Laine, has no shooter history so the model assumes him to be a league average finishing talent.

Tyler Myers is the Jets only over performing defender, with one goal scored more than we would expect. The rest of the Jets’ defenders have under performed, with Paul Postma leading the pack.

Closing Thoughts

As it currently stands, the Jets’ are winning games like a team out of the playoffs and the underlying numbers suggest that things will continue this way unless they have some dramatic changes.

The Jets have some players over performing, while others are underperforming. For the most part these have canceled out, as the team has been expected to score 57.6 5v5 goals and scored 58. Goaltending has been a bit of a struggle, with the team expected to allow 56.4 5v5 goals and allowed 60.