Biggest Regression Candidates

Mark Barbour (@18sktrs,

In this article I’m going to talk about the skaters who have scored either too many or too few goals during the last 20 team games. WARNING: I’ve got bad news about one of your favourite skaters (probably). I also have some good news. Let’s get into it.

Regression – What Are We Talking About Here?

I want to be clear about what I mean by “regression” in this article. I’m using that term to refer to the output of a linear regression model. The “predicted” goals that I talk about were predicted by that model.

With that out of the way, let’s look at the candidates for “positive” and “negative” regression. Note that only skaters who play at the forward positions are included in this analysis.

Biggest Over/Under Goal Scorers

The plot below is based on the difference between a skater’s actual goals and his predicted goals. The line represents the density of skaters at each point between -4 goals under predicted to +4 goals over predicted. Most skaters are clustered around “0.0” which means their actual goals are close to their predicted goals.

I highlighted 3 skaters at each end of the distribution – the skaters on the left scored fewer goals than predicted and the skaters on the right scored more goals than predicted.

Let’s start with the skaters who have scored too many goals and could see some “negative regression”.

  • I’ve talked about individual expected goal differential (or “ixGd”) in previous articles. Basically it compares a skater’s actual goals to his expected goals. An ixGd above 100% means a skater has scored more goals than expected and vice versa. Some skaters can maintain ixGd rates of 130-150% or even a little higher. Why do I mention that here? Jared McCann has an ixGd of 422% over the last 20 games. Over the last 10 games its 626%. That is not sustainable. If you can find someone who will give up valuable assets for McCann based on his recent run then go for it.
  • It seems everyone is in love with Tage Thompson these days. Hard to blame them given that Thompson has scored 18 goals in the last 20 games. Was that a few too many? Probably, but his “regressed” goals are still among the top skaters in the league (trailing only 4 skaters, including Jack Hughes). There’s probably not much actionable advice here. Tage Thompson has been terrific and all signs point to that continuing. If you can somehow find a Thompson-skeptic then I suggest you take advantage of the situation.
  • Maybe Klim Kostin has some value in deeper leagues (or leagues that reward hits heavily) and you’re trying to assess his recent hot streak. The regression model gives Kostin credit for scoring about 5 or 6 goals over the last 20 games. Set your expectations accordingly.

Now on to the “positive regression”!

  • Brady Tkachuk never scores as many goals as my regression models predict. His shooting stats are unique and it messes with the calculation. I’m actually in the process of building my own “expected goals” model (not a regression model like the one discussed here) that will properly account for an outlier like Brady Tkachuk. In the meantime we can observe that Tkachuk’s rate stats have been well below his “normal” during the last 20 games. His Goals/60 during that time was roughly half the rate he’s put up over the last 80 games. His xGoals/60 during the last 20 games was more than three standard deviations to the upside while his actual Goals/60 was slightly negative. That’s pretty nuts. He’s a buy low candidate though one that will probably be difficult to acquire.
  • Andrei Svechnikov is another skater who seems to always perform below expectations. As was the case with Tkachuck we can take that into account and still observe that Svechnikov’s recent performance has been below his “normal”. His stats like xGoals/60, Shots/60 and iSCF/60 have been trending up while his actual Goals/60 have been trending down. His ixGd over the last 20 games is 60% while is 80-game ixGd is 96%. He’s also a buy low candidate, and he’s probably a little easier to acquire than Tkachuk.
  • John Tavares has been putting up some excellent shooting stats over the last 20 games and he should probably have a couple more goals scored during that time. The problem is that he hasn’t been in the kind of funk that would have people tempted to get him off the roster. I doubt there’s much opportunity here.

The End Of The Article

That’s all for this week. I’ll be back here next week with another original data-driven article. In the meantime you can find my daily posts on Consider following me on Twitter at @18sktrs if you want to be notified about the content I’m creating here, there and anywhere.


Mark (18 Skaters)

The Data

The data for this article were pulled from Natural Stat Trick and include games played on or before 2023-01-17.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: