Score-Adjusted Fenwick standings update

WINNIPEG, CANADA - FEBRUARY 21: Jaromir Jagr #68 of the Philadelphia Flyers celebrates with Jakub Voracek #93 and Andrej Meszaros #41 after learning that Score-Adjusted Fenwick suggests the Flyers are better than what their Fenwick Tied had previously implied. (Photo by Tom Szczerbowski/Getty Images)


Shot differential (Corsi or Fenwick) has been shown to correlate strongly to puck possession and zone time, and to be a good predictor of future goal differential and wins. Previously, we showed how adjustments for score effects can improve the predictions appreciably and gave the Score-Adjusted Fenwick (SAF) standings.

It's been a month. After the jump, we'll update those standings and look at what's changed.

Here's the big table of raw data; at the bottom are some observations.

Team Score-Adjusted Fenwick Prev SAF Fenwick Tied Prev Fen Tied Fenwick Close Prev Fen Close
DET 56.1 (1) 57.0 (1) 56.6 (2) 56.5 (2) 56.2 (1) 56.8 (1)
STL 55.2 (2) 56.2 (2) 57.3 (1) 56.5 (1) 56.1 (2) 56.1 (2)
PIT 54.9 (3) 55.5 (3) 54.6 (3) 54.9 (3) 54.5 (3) 55.4 (3)
SJ 53.2 (4) 52.3 (7) 52.9 (6) 51.6 (8) 52.3 (6) 51.1 (9)
BOS 53.1 (5) 52.9 (6) 53.5 (4) 52.4 (5) 52.9 (4) 52.7 (4)
CHI 53.1 (6) 53.0 (4) 53.1 (5) 53.9 (4) 52.3 (5) 52.7 (5)
VAN 52.7 (7) 53.0 (5) 50.2 (11) 50.8 (11) 51.0 (10) 51.4 (6)
PHI 51.6 (8) 51.1 (8) 49.8 (16) 48.0 (23) 52.0 (7) 50.9 (11)
LA 51.0 (9) 50.9 (9) 52.2 (7) 51.8 (7) 51.2 (8) 51.3 (7)
WPG 50.7 (10) 50.8 (10) 49.9 (13) 50.4 (14) 50.8 (11) 51.1 (10)
NJ 50.2 (11) 49.7 (13) 51.7 (8) 51.3 (9) 51.1 (9) 50.3 (13)
OTT 50.2 (12) 49.5 (14) 50.1 (12) 50.4 (13) 50.2 (14) 50.1 (14)
COL 49.9 (13) 50.3 (11) 48.4 (20) 50.2 (15) 50.4 (12) 51.3 (8)
PHX 49.7 (14) 49.0 (18) 49.9 (15) 49.4 (18) 50.3 (13) 49.8 (16)
DAL 49.6 (15) 50.0 (12) 49.9 (14) 50.0 (16) 49.2 (18) 49.0 (19)
NYR 49.1 (16) 47.8 (24) 49.8 (17) 48.5 (21) 49.4 (16) 48.5 (21)
MTL 49.0 (17) 49.2 (16) 47.3 (26) 47.8 (24) 48.2 (24) 48.6 (20)
FLA 48.6 (18) 48.6 (19) 51.2 (9) 50.8 (10) 49.6 (15) 49.7 (17)
NYI 48.6 (19) 49.0 (17) 48.7 (18) 50.0 (17) 48.7 (21) 49.4 (18)
TOR 48.5 (20) 48.5 (20) 47.9 (23) 48.1 (22) 48.3 (23) 48.4 (22)
BUF 48.4 (21) 47.6 (26) 47.4 (25) 47.1 (26) 47.5 (25) 46.8 (27)
CAR 48.3 (22) 48.1 (23) 48.4 (19) 47.3 (25) 48.9 (19) 48.1 (24)
WSH 48.0 (23) 48.3 (22) 50.9 (10) 52.2 (6) 49.2 (17) 50.0 (15)
TB 47.9 (24) 47.8 (25) 47.1 (27) 46.7 (27) 48.8 (20) 48.1 (25)
CBJ 47.6 (25) 49.2 (15) 48.1 (21) 50.6 (12) 48.5 (22) 50.6 (12)
ANA 47.4 (26) 46.3 (28) 47.0 (28) 46.0 (28) 46.5 (28) 44.9 (30)
EDM 47.2 (27) 48.3 (21) 48.1 (22) 49.0 (19) 47.3 (26) 48.3 (23)
CGY 46.8 (28) 47.6 (27) 47.5 (24) 49.0 (20) 46.9 (27) 47.8 (26)
NSH 45.7 (29) 46.0 (29) 44.5 (30) 44.0 (29) 45.1 (30) 45.2 (28)
MIN 45.1 (30) 44.2 (30) 45.7 (29) 43.7 (30) 45.9 (29) 45.0 (29)

The fastest riser was the Rangers, who had a SAF of 47.8% through 46 games and are now at 49.1% through 58 games, suggesting their SAF over the last month was about 54.1%. Other risers include Anaheim (+1.1%), Minnesota (+0.9%), and San Jose (+0.9%).

The biggest drop was from the Blue Jackets, who had a SAF of 49.2% through 47 games and are now at 47.6% through 60 games, suggesting their SAF over the last month was about 41.8%. Other fallers include Edmonton (-1.1%), St. Louis (-1.0%), and Detroit (-0.9%).

Teams which SAF views more favorably than Fenwick Tied does include Vancouver (52.7% vs 50.2%), Philadelphia (51.6% vs 49.8%), Montreal (49.0% vs 47.3%), and Colorado (49.9% vs 48.4%).

Teams which SAF views less favorably than Fenwick Tied does include Washington (48.0% vs 50.9%), Florida (48.6% vs 51.2%), St. Louis (55.2% vs 57.3%), and New Jersey (50.2% vs 51.7%).

Star_divide_medium

When this methodology was published, we took an extensive look at various metrics to show that the larger sample size of SAF gives a more repeatable measure of talent (higher split-half reliability) and gives better predictions (better correlations to future points%), especially over the first half of the year. However, some dismissed our score adjustments based on the observation that the two metrics end up similar by the end of the year. So let's look at whether -- and how -- that harmonizing is occurring this year.

The two metrics do appear to be getting closer together. League-wide, the coefficient of determination (r^2) between SAF and Fenwick Tied increased from 0.80 to 0.83 over the last month. Of the ten teams that had a gap of at least 1% between the metrics a month ago, none got appreciably bigger (max increase was 0.3%), but five got quite a bit smaller (decrease between 0.7% and 1.3%). The two metrics seem to be converging over time.

Next, let's look at how the metrics are converging. For the five teams whose gap between SAF and Fenwick Tied narrowed considerably in the last month, we can check whether the two metrics are moving towards each other to meet in the middle or if one of them is changing to catch up to the other.

  • For one team -- Nashville -- the two metrics moved towards each other. Nashville's Fenwick Tied went up by 0.5% and their SAF went down by 0.3% as the metrics met in the middle.
  • For four teams -- Philadelphia, Washington, Columbus, and Calgary -- their Fenwick Tied moved sharply towards their SAF.
  • No team saw their SAF move sharply towards their Fenwick Tied.

Thus, there were four teams for which SAF told us a month ago what Fenwick Tied is just starting to suggest now, and there are no teams for which Fenwick Tied was ahead of SAF.

This is precisely what we would expect the benefit of the larger sample to be: SAF gives a more accurate talent evaluation earlier in the year. This is true across the board, and not just for these five teams -- as would be expected from the greater split-half reliability of SAF, the average change in Fenwick Tied was nearly twice as large as for SAF (0.9% vs 0.5%).

At this point in the year, Score-Adjusted Fenwick has largely stabilized at levels reflective of team talent, while SAF Fenwick Tied is still catching up. By the end of the season, the two metrics may give similar numbers, but for in-season predictions, the greater sample size is important. Moreover, given the enhanced importance of post-deadline performance, even end-of-year playoff predictions may benefit from having the kind of sample size that permits one to make good predictions while focusing on ~20-30 games of data.

X
Log In Sign Up

forgot?
Log In Sign Up

Forgot password?

We'll email you a reset link.

If you signed up using a 3rd party account like Facebook or Twitter, please login with it instead.

Forgot password?

Try another email?

Almost done,

Join Broad Street Hockey

You must be a member of Broad Street Hockey to participate.

We have our own Community Guidelines at Broad Street Hockey. You should read them.

Join Broad Street Hockey

You must be a member of Broad Street Hockey to participate.

We have our own Community Guidelines at Broad Street Hockey. You should read them.

Spinner

Authenticating

Great!

Choose an available username to complete sign up.

In order to provide our users with a better overall experience, we ask for more information from Facebook when using it to login so that we can learn more about our audience and provide you with the best possible experience. We do not store specific user data and the sharing of it is not required to login with Facebook.

tracking_pixel_9355_tracker