NHL Advanced Stats: Expected vs Observed Goal Differentials

 

A useful advanced stats for evaluating NHL team performance

With War-on-Ice set to cease operations by the end of the month, NHL stats fans are fortunate that Emmanuel Perry has created Corsica. The new advanced stats hub has most of the features that viewers came to rely on at WOI. But what sets Manny’s site apart is his inclusion of a variety of new stats. This kind of data allows for NHL analysis to drive a little deeper and uncover a little more understanding than we’ve had before.

One of the most interesting stats Manny has included is expected goals data. Follow the link to read up on Manny’s method (and turn off your ad blocker before you click, please!). For our purposes, here’s the snippet we need:

manny

Accounting for shot type, distance, and angle, rebounds, rush shots, and strength state, all shots are assigned a “goal expectancy.” These expected goal stats (xG) can then be applied to players and teams in a variety of ways.

So, here’s what I’ve made with the data.

 

gd60

 

At the team level, Corsica offers observed goals for per 60 (GF60) and goals against per 60 (GA60). To create a goals differential per 60, I’ve subtracted GA60 from GF60. That’s marked by a small dot for each team in the graph above.

Corsica’s xG data is offered at the team level as well. To create the expected goals differential per 60 (xGD60), I’ve repeated the process from above – expected goals against per 60 (xGA60) is subtracted from goals for per 60 (xGF60). On the graph above, the xGD60 is marked by the team label.

 

Reading the graph

For some teams, the team label is on the top of the team’s line. That means that the team’s xGD60 is greater than their actual GD60 this season – in other words, the team should have a better goals differential than they actually have this season.

If the team label is at the bottom of the line, the opposite is true – the team has enjoyed a goal differential better than we would expect.

The length of the line gives an indication of how great the differential is between observed and expected results. For the Leafs, there’s a huge gap between their poor observed GD60 and their xGD60. Toronto should have performed better this year based on shot quality.

For Chicago, there’s almost no difference between observed and expected results. They’ve achieved almost exactly the differential that Corsica’s shot quality model would predict.

 

Some thoughts on teams

 

No team has enjoyed exceeded expectations more than the New York Rangers. They boast a ~0.6 GD60 at 5v5, good for second best in the league behind only the Washington Capitals.

However, based on the xG data, NYR should actually have a mark closer to ~-0.3. That’s almost a full goal per 60 drop based on shot quality expectations.

Conversely, the Toronto Maple Leafs suffer the largest expectation deficit. The Leafs have registered a ~0.5 GD60 at 5v5 this season but, based on xG data, should likely be closer to a postive GD60 of ~0.2. That move from a negative to a positive differential would likely make for a very different wins-loss record.

Whether or not Leafs fans would actually prefer a few more wins (at the expense of a strong chance at drafting Auston Matthews) remains in doubt.


 

Read more…

NHL Game Charts – Sunday, March 20

The Protected List: Crowdsourcing Sens Twitter

A look back at the PDO Range in the NHL

 

Questions? Comments? Complaints? Contact me – seantierney1 AT hotmail

 

 

 

Advertisements

7 thoughts on “NHL Advanced Stats: Expected vs Observed Goal Differentials

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s