Re-visiting the relationship between Corsi and team standings

Around the halfway point of the 2015-2016 NHL season (on January 9th, to be exact), I looked into the relationship between 5v5 score adjusted Corsi For percentage and team standings points.

I found a couple of things. First off, “from 2007-2015, the R^2 between CF% and standings points is .4, which isn’t exactly incredibly high, but is still noteworthy.”

shane corsi 1

Also, up to that point of the 2015-2016 season, the relationship between the two was much weaker, with the R^2 between the two only being .18.

shane corsi 2

At the end of the article, I concluded that it would be interesting to revisit these results at the end of the season, in order to see what effects regression had on the standings. Up to that point in the year, there were five teams with a PDO over 101.0, and a CF% under 49.5%, as well as five teams with a CF% over 51.0%, but a PDO under 99.5.

Here are the teams that had high PDO, but low CF%.

Team CF% PDO
New York Rangers 48.2% 102.7
Florida Panthers 48.1% 102.4
Minnesota Wild 49.1% 101.6
Arizona Coyotes 47.3% 101.4
Ottawa Senators 46.1% 101.3
Colorado Avalanche 44.6% 101.1

And here are the teams that had high CF%, but low PDO.

Team CF% PDO
Anaheim Ducks 53.3% 96.9
Carolina Hurricanes 53.2% 97.5
Nashville Predators 54.2% 98.8
Chicago Blackhawks 52.1% 99.2
St. Louis Blues 53.0% 99.3

I also predicted that there would be some change in the standings at the end of the year, mainly due to regression.

So how did the relationship between CF% and the standings look at the end of the season?

CF% vs Standings 2015-2016

As we can see, it strengthened. The R^2 between CF% and standings points in 2015-2016 was 0.36, which is much higher than the .18 we saw at the halfway point of the season.

As expected, a large part of this can be explained by PDO regression. By the end of the year, only four teams had a PDO over 101.0, and only three teams had a PDO below 99.0. The clustering of PDO between 99 and 101 meant that CF% had a stronger influence over standings points, due to random variance playing a lesser role in team results.

At the same time, the 2015-2016 season continues the trend of declining returns for shot attempt metrics. A large reason for their rise to prominence was that they correlated well with standings points, but over the past couple of seasons, the relationship has been declining. Here’s a year-by-year look at the R^2 between CF% and standings points (lockout-shortened season omitted).

R^2

It will be interesting to see whether or not this trend continues in the future. Of course, if every team in the league only considered shot attempt metrics in roster construction, other aspects of the game (goaltending, special teams, shooting talent) would likely end up being the difference makers among teams.

With teams such as the Carolina Hurricanes posting high CF% numbers and low PDO numbers (due to both terrible goaltending and a dearth of shooting talent), and other teams such as the New York Rangers riding strong goaltending and talented shooters to the top of the standings despite low CF% numbers, it’s likely that we’ll see the trend of diminishing returns continue in 2016-2017. What the future holds after that, however, is unclear.

2 thoughts on “Re-visiting the relationship between Corsi and team standings

  1. Pingback: Weekly Sports Analytics News Roundup - September 6, 2016 - StatSheetStuffer

  2. Hi Shane,

    Would be curious to see whether or not the coefficients on CF% were significant, since they’re so big – looks like a 1% increase in CF% leads to 2.5-2.75 more points.

    Thanks,

    Gord

Leave a Reply

Your email address will not be published. Required fields are marked *

*