FanPost

Why QOC Doesn't Matter

QOC doesn't seem to impact production as measured by Corsi . Corsi QOC is based on the premise that teams produce Corsi events at a consistent rate. However, teams and players do not produce Corsi events at a steady rate. Just because we know that a team has an average rate of, say, +4, what might they do in any particular game? To what extent does QOC actually predict the Quality of Competition? Does knowing the overall average Corsi rate help much? The answer is: not really. I took the 2011-2012 data, and looked at each team game by game. I generated the mean, standard deviation, maximum, and minimum values for CorsiTeam.

team

mean

sd

Min

Max

ANA

-2.8527853

17.26377

-53.23475

39.75628

BOS

8.6254764

17.66912

-41.71963

49.98386

BUF

-2.5102399

17.03829

-40.571

33.17479

CAR

-3.8438323

18.37554

-43.72948

38.62928

CBJ

-5.3158449

18.504

-52.76596

36.74858

CGY

-5.3088753

17.88541

-48.63039

42.79835

CHI

6.4708799

14.62719

-33.8961

47.23524

COL

3.0242469

18.46124

-48.31288

42.37013

DAL

-0.5312207

16.5221

-39.40008

39.19129

DET

10.0239904

14.77338

-15.34289

53.23475

EDM

-6.6591129

19.98099

-59.75104

28.29341

FLA

-0.8878748

18.15345

-39.13043

38.91167

L.A

10.7316018

18.7189

-30.40113

67.82609

MIN

-12.2736013

19.02055

-67.82609

36.95575

MTL

-4.7486317

18.38171

-71.74377

29.64997

N.J

0.1662205

18.57297

-35.13419

48.63039

NSH

-7.2526589

17.08839

-49.4889

46.34146

NYI

-5.5527745

15.39438

-36.80424

30.63263

NYR

-5.2923161

16.9775

-49.98386

34.8116

OTT

5.7136312

19.24359

-38.69626

71.74377

PHI

2.3509001

19.21295

-44.47987

45.65217

PHX

-0.3857895

18.83309

-45.2188

46.73727

PIT

9.4548142

16.94484

-27.68263

43.72948

S.J

4.10164

19.20484

-47.23524

45.2188

STL

6.959637

18.21388

-29.53978

52.76596

T.B

-6.0466816

19.94139

-49.27007

34.18468

TOR

-2.6417938

19.14135

-49.58968

35.28176

VAN

5.7467211

18.61605

-37.76824

46.7706

WPG

0.7512969

15.54967

-36.85504

39.13043

WSH

-1.9209396

19.67149

-38.62928

59.75104

The average standard deviation here is just under 18, which gives a 90% confidence interval of about +/- 30. The results are similar in previous years. If we look at the results of the Blues versus the Eastern Conference teams in 2011 we see:

Team

Game

Corsi

BOS

20897

26.7

BUF

20707

6.12

CAR

20093

14.1

CAR

21055

26.69

FLA

20264

1.41

MTL

20622

1.38

N.J

20801

2.35

NYI

20854

1.31

NYR

20456

11.43

OTT

20791

-27.58

PHI

20100

-2.44

PIT

20303

13.12

PIT

20728

-4.06

T.B

20236

28.08

T.B

21071

-21.06

TOR

20216

32.75

WPG

20915

-26.85

WSH

20347

24.21

It can be argued that using the season average Corsi to predict performance has a bias by including games against that opponent. While true, the effect is small. Looking at the Blues games against their Central Division rivals, we can adjust their Corsi by removing the games against each opponent. I have included the Blues maximum, minimum, and average Corsi output in the 6 games against each opponent.

Opponent

Blues Corsi

Adj Corsi

Max

Min

Average

CBJ

6.95

6.71

52.77

-11.75

10.03

CHI

6.95

7.62

17.53

-19.84

-1.39

DET

6.95

7.93

12.18

-10.92

-5.28

NSH

6.95

6.62

36.00

-15.02

11.19

So knowing the season average Corsi of the Blues tells us little about how they will perform on any given night. This is true across the board. Given that many teams only face each other once or twice and all teams face each other at most a handful of times, it is easy to see why Corsi QOC has little bearing on observed performance.

If the standard deviation for one game is about 18, the standard deviation for the average of 4 games is almost 9, and the standard deviation for the average of 6 games is about 7.3. Given that the best team in the league is usually around +10, and the worst around -10, even with 6 games against an average (Corsi=0) team, their average production for those 6 games could be better than the expected value for the best team in the league, or worse than the expected value for the worst team in the league. If we wanted the standard deviation for the average to be less than 5, teams would have to play each other 13 times. If we wanted the standard deviation for the average to be less than 3, teams would have to play each other 36 times.

Corsi QOC Predictive Value – Players

Compared to Team Corsi, Player Corsi is going to be even more variable game to game. Looking at the Blues who played more than 10 games in 2012 and analyzing the game by game Corsi data, we see

name

Corsi

sd

Corsi Rel

sd

n

BACKES

4.78

27.27

1.36

27.24

48

BERGLUND

-2.31

26.38

-8.11

33.49

48

BOUWMEESTER

9.76

26.57

7.63

29.73

14

COLE

3.56

27.22

3.69

30.06

15

CRACKNELL

7.81

42.81

9.04

44.98

20

D'AGOSTINI

-0.94

30.25

-0.77

27.88

16

JACKMAN

-7.16

24.53

-15.39

29.46

46

LEOPOLD

12.03

27.20

9.35

29.36

15

MCDONALD

3.00

28.17

-4.29

32.05

37

NICHOL

-0.68

32.23

-3.66

30.27

30

OSHIE

-0.35

24.94

-4.91

23.08

30

PERRON

4.84

24.48

1.62

25.07

48

PIETRANGELO

4.81

26.60

0.66

30.79

47

POLAK

-0.96

27.35

-6.85

28.69

48

PORTER

3.08

35.88

-2.39

38.86

29

REAVES

1.18

31.63

-0.69

35.79

43

REDDEN

14.13

26.61

14.29

30.43

23

RUSSELL

-0.53

30.99

-5.61

28.61

33

SCHWARTZ

1.29

30.43

-3.62

30.51

45

SHATTENKIRK

10.55

24.83

9.86

28.61

48

SOBOTKA

7.91

27.39

5.74

26.47

48

STEEN

10.49

31.15

6.83

30.42

40

STEWART

-3.49

27.04

-9.95

29.47

48

TARASENKO

15.46

30.28

12.64

31.23

38

Here the average standard deviation of Corsi is almost 28 and the average standard deviation of Corsi Rel is almost 31. Knowing that David Backes is +4.78 really only tells you that on any given night there is a 90% probability that his performance will be between -41.03 and +50.59. That is obviously not very helpful.

Looking at samples of size 4 and 6, the standard deviation for the average of 4 games is about 14, and the standard deviation for the average of 6 games is about 11.4. Not enough to allow an average player to look like the best player in the league, but enough that the 90% confidence interval of the average of 6 games is about +/- 19.

Corsi Rel QOC Predictive Value – Players

In theory, Corsi Rel should, in essence, rank-order players, and Corsi Rel QOC should, in essence, rank-order the opposition. The variability Corsi and Corsi Rel are enough that players with low averages could, in any given game, have a higher Corsi than players with high averages. So I looked at the game by game Corsi Rel of pairs of players. First Barrett Jackman and Alex Pietrangelo. If you look at the difference in their Corsi Rel, while Pietrangelo tends to be better, the results are all over the map. While the mean difference is 15.26, the standard deviation of that difference is 29.05. In fact, in 15 or their 44 games, Jackman's Corsi Rel was higher than Pietrangelo's. This variability is not unique to these two players but is true across the board. Looking at David Backes and Patrik Berglund, the difference is 6.62 but the standard deviation of the difference is 36.65.

Looking specifically at Backes, his Corsi Rel of 1.36 only tells you that on any given night there is a 90% probability that his performance will be between -44.40 and +47.12. Again, not helpful.

Also, once again 4 game and 6 game averages fail to reduce the standard deviation enough to be helpful.

Just as Corsi QOC fails to reliably predict the actual Quality of Competition for any given game, Corsi Rel QOC fails to reliably predict the relative difficulty of the opposition.

Conclusions

Corsi rates are not fixed. They are highly variable from game to game. Since each team only plays each other team a few times in a season, Corsi QOC fails to reliably predict the actual Quality of Competition at either a team level or a player level. Corsi Rel QOC fails to reliably predict the relative difficulty of the opposition players.

Please make sure that any content you post is appropriate to Game Time, which means that it pertains to hockey, the Blues, frosty adult beverages, or puppies.

X
Log In Sign Up

forgot?
Log In Sign Up

Please choose a new SB Nation username and password

As part of the new SB Nation launch, prior users will need to choose a permanent username, along with a new password.

Your username will be used to login to SB Nation going forward.

I already have a Vox Media account!

Verify Vox Media account

Please login to your Vox Media account. This account will be linked to your previously existing Eater account.

Please choose a new SB Nation username and password

As part of the new SB Nation launch, prior MT authors will need to choose a new username and password.

Your username will be used to login to SB Nation going forward.

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,

By becoming a registered user, you are also agreeing to our Terms and confirming that you have read our Privacy Policy.

Join St. Louis Game Time

You must be a member of St. Louis Game Time to participate.

We have our own Community Guidelines at St. Louis Game Time. You should read them.

Join St. Louis Game Time

You must be a member of St. Louis Game Time to participate.

We have our own Community Guidelines at St. Louis Game Time. You should read them.

Spinner.vc97ec6e

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